Why spatial modelling Spatial data structure Software and data
Spatial EconometricsLecture 1 The notion of spatial modelling Visualisation of
spatial data in R
Andrzej Toroacutej
Institute of Econometrics Department of Applied Econometrics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 1 30
Why spatial modelling Spatial data structure Software and data
Contents
1 Why spatial modelling
Space vs ties between units
Applications of spatial modelling
2 Spatial data structure
Spatial order vs temporal order
Types of spatial inuences
3 Software and data
Software
Sources of spatial data and literature
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 2 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 3 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Paradigms
Toblers law (1970) Everything is related to everything else
but near things are more related than distant things
Toblers second law (2004) The word near may have a lot
of meanings
In other words Beck (2006) There is more to space than
geography
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 4 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Paradigms
Toblers law (1970) Everything is related to everything else
but near things are more related than distant things
Toblers second law (2004) The word near may have a lot
of meanings
In other words Beck (2006) There is more to space than
geography
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 4 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Paradigms
Toblers law (1970) Everything is related to everything else
but near things are more related than distant things
Toblers second law (2004) The word near may have a lot
of meanings
In other words Beck (2006) There is more to space than
geography
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 4 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Case 1 neighbourhood ties between counties in USA
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 5 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Case 2 following tweets among French members ofparliament
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 6 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
How are spatial ties generated
Units impact on one anothereg spatial diusion cases of u in a given district
But also
Level of measurement not suitable for the investigatedphenomenon (np regional aggregates instead of micro-data)Common measurement errors (eg community-level dataprepared under dierent guidelines of regional statisticaloces)Spatial aggregation level
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 7 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (1)
On average 0 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 8 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (2)
On average 267 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 9 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (2)
security policy (alliances wars military interventions)
protection of environment (air pollution water contagion)
international interdependence of policies (copying legislation
patterns)
political science (constituencies and electoral systems)
epidemiology (spreading of epidemics)
economic diusion (local labour markets)
More Haegerstrand (1967) Manski (2000) Simmons et al(2005)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 11 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 12 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Contents
1 Why spatial modelling
Space vs ties between units
Applications of spatial modelling
2 Spatial data structure
Spatial order vs temporal order
Types of spatial inuences
3 Software and data
Software
Sources of spatial data and literature
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 2 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 3 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Paradigms
Toblers law (1970) Everything is related to everything else
but near things are more related than distant things
Toblers second law (2004) The word near may have a lot
of meanings
In other words Beck (2006) There is more to space than
geography
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 4 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Paradigms
Toblers law (1970) Everything is related to everything else
but near things are more related than distant things
Toblers second law (2004) The word near may have a lot
of meanings
In other words Beck (2006) There is more to space than
geography
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 4 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Paradigms
Toblers law (1970) Everything is related to everything else
but near things are more related than distant things
Toblers second law (2004) The word near may have a lot
of meanings
In other words Beck (2006) There is more to space than
geography
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 4 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Case 1 neighbourhood ties between counties in USA
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 5 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Case 2 following tweets among French members ofparliament
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 6 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
How are spatial ties generated
Units impact on one anothereg spatial diusion cases of u in a given district
But also
Level of measurement not suitable for the investigatedphenomenon (np regional aggregates instead of micro-data)Common measurement errors (eg community-level dataprepared under dierent guidelines of regional statisticaloces)Spatial aggregation level
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 7 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (1)
On average 0 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 8 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (2)
On average 267 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 9 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (2)
security policy (alliances wars military interventions)
protection of environment (air pollution water contagion)
international interdependence of policies (copying legislation
patterns)
political science (constituencies and electoral systems)
epidemiology (spreading of epidemics)
economic diusion (local labour markets)
More Haegerstrand (1967) Manski (2000) Simmons et al(2005)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 11 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 12 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 3 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Paradigms
Toblers law (1970) Everything is related to everything else
but near things are more related than distant things
Toblers second law (2004) The word near may have a lot
of meanings
In other words Beck (2006) There is more to space than
geography
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 4 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Paradigms
Toblers law (1970) Everything is related to everything else
but near things are more related than distant things
Toblers second law (2004) The word near may have a lot
of meanings
In other words Beck (2006) There is more to space than
geography
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 4 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Paradigms
Toblers law (1970) Everything is related to everything else
but near things are more related than distant things
Toblers second law (2004) The word near may have a lot
of meanings
In other words Beck (2006) There is more to space than
geography
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 4 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Case 1 neighbourhood ties between counties in USA
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 5 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Case 2 following tweets among French members ofparliament
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 6 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
How are spatial ties generated
Units impact on one anothereg spatial diusion cases of u in a given district
But also
Level of measurement not suitable for the investigatedphenomenon (np regional aggregates instead of micro-data)Common measurement errors (eg community-level dataprepared under dierent guidelines of regional statisticaloces)Spatial aggregation level
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 7 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (1)
On average 0 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 8 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (2)
On average 267 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 9 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (2)
security policy (alliances wars military interventions)
protection of environment (air pollution water contagion)
international interdependence of policies (copying legislation
patterns)
political science (constituencies and electoral systems)
epidemiology (spreading of epidemics)
economic diusion (local labour markets)
More Haegerstrand (1967) Manski (2000) Simmons et al(2005)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 11 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 12 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Paradigms
Toblers law (1970) Everything is related to everything else
but near things are more related than distant things
Toblers second law (2004) The word near may have a lot
of meanings
In other words Beck (2006) There is more to space than
geography
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 4 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Paradigms
Toblers law (1970) Everything is related to everything else
but near things are more related than distant things
Toblers second law (2004) The word near may have a lot
of meanings
In other words Beck (2006) There is more to space than
geography
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 4 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Paradigms
Toblers law (1970) Everything is related to everything else
but near things are more related than distant things
Toblers second law (2004) The word near may have a lot
of meanings
In other words Beck (2006) There is more to space than
geography
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 4 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Case 1 neighbourhood ties between counties in USA
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 5 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Case 2 following tweets among French members ofparliament
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 6 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
How are spatial ties generated
Units impact on one anothereg spatial diusion cases of u in a given district
But also
Level of measurement not suitable for the investigatedphenomenon (np regional aggregates instead of micro-data)Common measurement errors (eg community-level dataprepared under dierent guidelines of regional statisticaloces)Spatial aggregation level
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 7 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (1)
On average 0 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 8 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (2)
On average 267 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 9 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (2)
security policy (alliances wars military interventions)
protection of environment (air pollution water contagion)
international interdependence of policies (copying legislation
patterns)
political science (constituencies and electoral systems)
epidemiology (spreading of epidemics)
economic diusion (local labour markets)
More Haegerstrand (1967) Manski (2000) Simmons et al(2005)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 11 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 12 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Paradigms
Toblers law (1970) Everything is related to everything else
but near things are more related than distant things
Toblers second law (2004) The word near may have a lot
of meanings
In other words Beck (2006) There is more to space than
geography
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 4 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Paradigms
Toblers law (1970) Everything is related to everything else
but near things are more related than distant things
Toblers second law (2004) The word near may have a lot
of meanings
In other words Beck (2006) There is more to space than
geography
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 4 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Case 1 neighbourhood ties between counties in USA
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 5 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Case 2 following tweets among French members ofparliament
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 6 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
How are spatial ties generated
Units impact on one anothereg spatial diusion cases of u in a given district
But also
Level of measurement not suitable for the investigatedphenomenon (np regional aggregates instead of micro-data)Common measurement errors (eg community-level dataprepared under dierent guidelines of regional statisticaloces)Spatial aggregation level
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 7 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (1)
On average 0 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 8 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (2)
On average 267 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 9 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (2)
security policy (alliances wars military interventions)
protection of environment (air pollution water contagion)
international interdependence of policies (copying legislation
patterns)
political science (constituencies and electoral systems)
epidemiology (spreading of epidemics)
economic diusion (local labour markets)
More Haegerstrand (1967) Manski (2000) Simmons et al(2005)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 11 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 12 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Paradigms
Toblers law (1970) Everything is related to everything else
but near things are more related than distant things
Toblers second law (2004) The word near may have a lot
of meanings
In other words Beck (2006) There is more to space than
geography
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 4 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Case 1 neighbourhood ties between counties in USA
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 5 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Case 2 following tweets among French members ofparliament
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 6 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
How are spatial ties generated
Units impact on one anothereg spatial diusion cases of u in a given district
But also
Level of measurement not suitable for the investigatedphenomenon (np regional aggregates instead of micro-data)Common measurement errors (eg community-level dataprepared under dierent guidelines of regional statisticaloces)Spatial aggregation level
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 7 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (1)
On average 0 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 8 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (2)
On average 267 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 9 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (2)
security policy (alliances wars military interventions)
protection of environment (air pollution water contagion)
international interdependence of policies (copying legislation
patterns)
political science (constituencies and electoral systems)
epidemiology (spreading of epidemics)
economic diusion (local labour markets)
More Haegerstrand (1967) Manski (2000) Simmons et al(2005)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 11 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 12 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Case 1 neighbourhood ties between counties in USA
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 5 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Case 2 following tweets among French members ofparliament
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 6 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
How are spatial ties generated
Units impact on one anothereg spatial diusion cases of u in a given district
But also
Level of measurement not suitable for the investigatedphenomenon (np regional aggregates instead of micro-data)Common measurement errors (eg community-level dataprepared under dierent guidelines of regional statisticaloces)Spatial aggregation level
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 7 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (1)
On average 0 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 8 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (2)
On average 267 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 9 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (2)
security policy (alliances wars military interventions)
protection of environment (air pollution water contagion)
international interdependence of policies (copying legislation
patterns)
political science (constituencies and electoral systems)
epidemiology (spreading of epidemics)
economic diusion (local labour markets)
More Haegerstrand (1967) Manski (2000) Simmons et al(2005)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 11 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 12 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Case 2 following tweets among French members ofparliament
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 6 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
How are spatial ties generated
Units impact on one anothereg spatial diusion cases of u in a given district
But also
Level of measurement not suitable for the investigatedphenomenon (np regional aggregates instead of micro-data)Common measurement errors (eg community-level dataprepared under dierent guidelines of regional statisticaloces)Spatial aggregation level
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 7 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (1)
On average 0 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 8 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (2)
On average 267 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 9 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (2)
security policy (alliances wars military interventions)
protection of environment (air pollution water contagion)
international interdependence of policies (copying legislation
patterns)
political science (constituencies and electoral systems)
epidemiology (spreading of epidemics)
economic diusion (local labour markets)
More Haegerstrand (1967) Manski (2000) Simmons et al(2005)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 11 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 12 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
How are spatial ties generated
Units impact on one anothereg spatial diusion cases of u in a given district
But also
Level of measurement not suitable for the investigatedphenomenon (np regional aggregates instead of micro-data)Common measurement errors (eg community-level dataprepared under dierent guidelines of regional statisticaloces)Spatial aggregation level
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 7 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (1)
On average 0 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 8 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (2)
On average 267 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 9 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (2)
security policy (alliances wars military interventions)
protection of environment (air pollution water contagion)
international interdependence of policies (copying legislation
patterns)
political science (constituencies and electoral systems)
epidemiology (spreading of epidemics)
economic diusion (local labour markets)
More Haegerstrand (1967) Manski (2000) Simmons et al(2005)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 11 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 12 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (1)
On average 0 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 8 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (2)
On average 267 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 9 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (2)
security policy (alliances wars military interventions)
protection of environment (air pollution water contagion)
international interdependence of policies (copying legislation
patterns)
political science (constituencies and electoral systems)
epidemiology (spreading of epidemics)
economic diusion (local labour markets)
More Haegerstrand (1967) Manski (2000) Simmons et al(2005)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 11 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 12 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Space vs ties between units
Spatial aggregation level (2)
On average 267 neighbours of the same colourAndrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 9 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (2)
security policy (alliances wars military interventions)
protection of environment (air pollution water contagion)
international interdependence of policies (copying legislation
patterns)
political science (constituencies and electoral systems)
epidemiology (spreading of epidemics)
economic diusion (local labour markets)
More Haegerstrand (1967) Manski (2000) Simmons et al(2005)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 11 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 12 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (2)
security policy (alliances wars military interventions)
protection of environment (air pollution water contagion)
international interdependence of policies (copying legislation
patterns)
political science (constituencies and electoral systems)
epidemiology (spreading of epidemics)
economic diusion (local labour markets)
More Haegerstrand (1967) Manski (2000) Simmons et al(2005)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 11 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 12 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (2)
security policy (alliances wars military interventions)
protection of environment (air pollution water contagion)
international interdependence of policies (copying legislation
patterns)
political science (constituencies and electoral systems)
epidemiology (spreading of epidemics)
economic diusion (local labour markets)
More Haegerstrand (1967) Manski (2000) Simmons et al(2005)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 11 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 12 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (1)
Applications
traditional maps of ties between states or regions eg intrade or foreign investmentregional analyses ties between relatively small units (poviatscommunities) observable via eg unemployment rates or localgovernment nancebusiness analyses big data from GIS systems (eg modellingthe attractiveness of business locations optimisation of salesnetwork managing logistics)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 10 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (2)
security policy (alliances wars military interventions)
protection of environment (air pollution water contagion)
international interdependence of policies (copying legislation
patterns)
political science (constituencies and electoral systems)
epidemiology (spreading of epidemics)
economic diusion (local labour markets)
More Haegerstrand (1967) Manski (2000) Simmons et al(2005)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 11 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 12 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Applications of spatial modelling
Applications of spatial modelling (2)
security policy (alliances wars military interventions)
protection of environment (air pollution water contagion)
international interdependence of policies (copying legislation
patterns)
political science (constituencies and electoral systems)
epidemiology (spreading of epidemics)
economic diusion (local labour markets)
More Haegerstrand (1967) Manski (2000) Simmons et al(2005)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 11 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 12 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 12 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Temporal order
For time series we use the notion of serial correlation It
makes sense when
observations are aligned in a linear order (1-D)the frequency of the series is set ie identically long reportingperiods (or intervals between measurement moments)
Source of information about the order the records are
sorted or there is an explicit timestamp
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 13 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Spatial order
The temporal 1-D order is not the only possible one for our
data
observations may be attributed to areas or points on a surfaceor sphere (2-D)implications of such an order are more dicult to manage (2-Drather than 1-D) but ignoring this may lead to the sameproblems as temporal serial correlation
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 14 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Source of information about the spatial order
In spatial econometrics the order is described by a spatial weight
matrix (see next lecture) It might be based on
manual imputation of neighbourhood relationships (tedious)
eg USA linked to Mexico Canada linked to USA Mexico notlinked to Canada
distance matrix between units in space
how to generate ithow exactly to measure the distance
graphical 2-D representation of space ie a map from which
neighbourhood relationships or distances can be derived
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 15 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Spatial order vs temporal order
Econometric implications of spatial linkages
Observations are not independent
In this model yi = β0 + β1xi + εi it is not any more true thatεi sim i i d (independent identically distributed)The consequence is at best ineciency of OLS estimation(like in the case of temporal autocorrelation)
The big dierence while the temporal impact runs in onedirection only (past rarr present) spatial autocorrelation canrun in both directions (our region larrrarr neighbourhoodregion larrrarr other regions larrrarr our region)
Here the implications can be more serious and involveinconsistency and bias in OLS estimation (like in simultaneousequations models)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 16 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences
will be presented by world-class experts in neighbourhood topics
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 17 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (1)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 18 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (2)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 19 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (3)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 20 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Main types of spatial inuences (4)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 21 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Types of spatial inuences
Types of spatial inuences additional remarks
Hasty conclusions of spatial interdependence should be avoided
(situation 3) unless other cases have reasonably been
excluded
Spatial interdepencence (situation 3) is relatively unlikely when
modelling spatial aggregates Situations 1 and 2 more likely in
that case
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 22 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Plan prezentacji
1 Why spatial modelling
2 Spatial data structure
3 Software and data
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 23 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Software
Software
A small number of econometric packages oers tools for spatial
modelling The leading ones Matlab Stata R The materials
accompanying this lecture use R (via RStudio)
Installing package spdep
installpackages(spdep)
library(spdep)
Another useful package is rgdal and for visualising data on map
additionally maptools RColorBrewer i classInt
Lista wszystkich pakietoacutew R do analizy danych przestrzennych na
CRAN
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 24 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example
Our task is to plot a map illustrating the unemployment rate from
the le BDL_danexls for poviats This sample covers Poland in
2014 and comes from Local Data Bank by GUS (Central
Statistical Oce in Poland)
To do this we must merge the unemployment data with
cartographic data
This is how we impose a spatial structure on the data
Will also be useful in modelling in the following lectures
Solution with comments in the accompanying R code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 25 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Example nal eect of the code
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 26 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Sources of cartographic data
GADM administrative divisions in almost all countries of the world
CODGiK more accurate maps by Polish Centralny Oplusmnrodek Dokumentacji
Geodezyjnej i Kartogracznej (surveyor authority)
Eurostat maps of EU states in NUTS nomenclature from NUTS0 toNUTS3 (in Poland NUTS2 voivodships NUTS4 poviats)
package cshapes in R ready-made world map (current and historicalafter 1945) with additional functions for international spatial analyses
and many more
Optimization of map size
Map les are usually ver precise taking many MB of disc space
One can decrease this size - often without any detriment to the
illustration quality - using converters (eg mapshaperorg)
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 27 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Functions in use
readOGR imports the map and uses the following les
shp shapes of the regionsadditionally shx (auxiliary indexing le) and dbf
(accompanying database)prj technical details related to the projection of geosphere onthe plane
spTransform allows to transform all the geocoding
information into longitude and latitude in degrees (required by
a number of R packages)
colorRampPalette useful to transform the visualised
(continuous) variable into colours
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 28 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Homework 1Using the Eurostat database and R please illustrate the regional variation in a selected variable in aEuropean country of Your choice and a chosen time period (this should not be the unemployment ratenor Poland but still a relatively large country)
One can use an own alternative data source The map should consist of 25-30 regions Theillustrated variable will be the dependend variable in the model prepared in subsequenthomeworks so one should be able to nd further variables for the same set of regions (to beused as sensible regressors)
The uploaded PDF le should contain a short descritption of the analysed variable data sourceand map source indication and the illustration (map) itself max 1 page
The uploaded ZIP le should contain the whole replication package data le R code andpossibly map les (if these are the same les as the ones used in the class this can be skipped ifthese are dierent les but prohibitively big ie making the upload size exceed the 5 MB limitplease put only a txt le with URL codes of every of the map les (or a ZIP package) on theuploaded archive This remark is related to all future homeworks
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 29 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30
Why spatial modelling Spatial data structure Software and data
Sources of spatial data and literature
Literature
Compulsory
Arbia G A Primer for Spatial Econometrics with Applications inR 2014 Palgrave Macmillan
Other
Anselin L Spatial Econometrics ch 29 in TC Mills and KPatterson (Eds) Palgrave Handbook of Econometrics Volume 1Econometric Theory Basingstoke Palgrave Macmillan 2006 pp901-969
Le Sage J and Pace RK Introduction to Spatial Econometrics2009 Chapman and HallCRC
Andrzej Toroacutej Institute of Econometrics Department of Applied Econometrics
(1) Spatial Econometrics 30 30