Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
茶の湯の名物記 ―『君台観左右帳記』の残照
Records of Distinguished Chanoyu Utensils ― an Afterglow of Kundaikansohchoki 同志社大学文化情報学部 矢野 環
CIS Doshisha University Tamaki YANO
A mathematical phylogenetic result about a cultural phenomenon is introduced. Formerly, we have inlaid imports from China as something valuable. "Kundaikan-Socho-Ki" was a standard text in Muromachi era when those chanoyu utensils are estimated,
KUNDAIKAN SOUCHOUKI 君台観左右帳記 A single-volume book detailing Chinese objects, particularly tea-ceremony utensils, given to the shogunal family by Ashikaga Yoshimasa's 足利義政 (1436-90) companion, the poet and artist, Nouami 能阿弥 (1397-1471). It is said that the author is either Nouami or Souami 相阿弥 (1455-1525). This book describes the decoration shoinkazari 書院飾り, of the formal reception rooms, for example at Yoshimasa's Higashiyama 東山 mountain villa, as well as describing Chinese paintings, incense containers, censers, flower vases, tea containers, trays, and giving contemporary appraisals of these items. http://www.aisf.or.jp/~jaanus/deta/k/kundaikansouchouki.htm
Left:Phylogeny right:Elliptic Fourier Analysis below:Famous Tea Pods
A quantitative approach for the Islamic urban structure Hiroomi TSUMURA
Islamic urban structure pointed out that the presence of legal provisions in its definition (Islamic law) is the most important. In addition, about the its space, it was pointed out that there has been such as separation of living space and public and private space by law Identification. In addition, Wirth (1982)2 and Ehler (1991)3, they were referring to the J.abu-Lughod's research thesis, and advancing the case studies of Islamic city. As a result, the Islamic
Haneda and Miura (1991)4 , the above generalized urban structure model was mentioned that not explain the each regional Islam urban structure individually. He insisted on the need for it to consider separately which the Islamic world can be divided into such as Maghreb (Western Arab), mashriq (Eastern Arab), Turkey, Persia and Central Asia. As J.abu-Lughod has argued, if the Islamic urban structure would be influenced on the legal provisions, his prospect is increasingly important. These discussions on the one hand unfortunately are conceptualized as a urban model, and they are described as the Ethnographies on the other hand.
The quo of the above Islamic urban structure researches, there is no mention of the coastal area of Arabian Sea and Mediterranean, in spite of another important "marine" Islamic world. From 2009, CKSCH was aimed to elucidate the "marine" Islamic urban structure, Oman (Sur), Egypt (Alexandria), Turkey (Istanbul), had been conducted ongoing researches. In this presentation, especially in those studies, would about the case of applying the spatial structure evaluation method using GIS is reported. Firstly introduces our field survey, and then will reported about the spatial data acquisition, and finally, quantitative analysis of those data - we will discuss network analysis.
1 JI . nAtebrun-aLtuiognhaold Jo 1u 9r 8n 7al oTf hMei Idsdllae mEica sCt itSyt -uHd ii se ts.o rViocl .1M9y-t2h. , Islamic Essence, and Contemporary Relevance.2 E.Wirth 1982 Villes Islamique,Villes Arabes, Villes Orientals?: Une problematique face au changement. La Ville 34 EM.. HEhalneerd 1a 9a9n1 d WTa.qMfi aurnad (tehde. ) C1i9ty9 1o . stT.o kFyaos cUicnl iev eb rilsl ii tnyg puree sd se . Arabe, Tunis and Paris.
f the Islamic Miccle Ea recherchehes. No.24.『イスラム都市研究』
①
Mosque or the other religious facilities are located in the center of city, In ②surrounding area of the center, the madrassas (religious schools) and the souks (markets) are arranged, General residential area is installed on the outside, and ③each facility are linked by the linear alley. Its planar shape will be concentric circle and the flow lines of residents would be radiation shape along the link.
urban structure, the following three characteristics were found to be typically.
Visual Statistical Learning Modulates Memory and Cognition Sachio OTSUKA
Doshisha University
Statistical learning is defined as the extraction of systematic regularities or patterns embedded in a continuous stream of stimuli (Saffran, Aslin, & Newport, 1996). In this study, I examined whether learning statistical regularities influences memory using a visual statistical learning paradigm.
Thirty-two students (14 men and 18 women, age range = 18–25 years) participated in an experiment that consisted of three phases: familiarization, recognition memory, and test. In the familiarization phase, participants observed a stream of objects. The stream consisted of structured triplets, where three objects were always presented in the same order, and random triplets. In the last part of the familiarization phase, Old Turkic letters were inserted into both structured and random triplets. The recognition memory phase was separated into two blocks (objects and Old Turkic letters). Each block contained 48 trials. Half of the 48 trials showed items presented in the familiarization phase, and the other half showed new items that were not previously presented. Participants were required to decide whether the items shown were old or new. Finally, to test whether participants learned the statistical regularities in temporal order during the first familiarization phase, I administered a two-alternative forced-choice familiarity test. In each test trial, participants judged the relative familiarity of two types of sequences based on the familiarization phase. One test sequence consisted of triplets that were presented in the stream. The other sequence was a foil constructed with items of three different triplets.
First, a one-sample t-test was conducted comparing the percentage of triplets judged as familiar in the test phase to chance (50%). This analysis showed that triplets were successfully discriminated from foils (55%), t(31) = 2.27, p = .030, indicating that participants learned the statistical regularities in temporal order in the familiarization phase. Recognition memory performance was measured as the percentage of tested items that were correctly identified as “old” or “new.” We performed a two-way within-subjects analysis of variance on percent correct in the recognition memory phase, with stimulus type (objects from the sequences vs. letters inserted into the sequences) and tested item type (structured vs. random) as factors. The results revealed a significant main effect of stimulus type, F(1, 31) = 53.11, p < .001, but the effect of tested item type was not significant, F(1, 31) = 0.05, p = .833. In addition, the interaction between stimulus type and tested item type was significant, F(1, 31) = 15.83, p < .001. The effect of tested item type was significant for both objects, F(1, 31) = 12.13, p = .002, and Old Turkic letters, F(1, 31) = 8.97, p = .005. Taken together, objects from structured triplets were more likely to be remembered than those from random triplets (81% vs. 74%), and letters inserted into the structured triplets were less likely to be remembered than were control letters (56% vs. 63%). These results suggest that visual statistical learning enhances memory for each element in a regular set and impairs memory for items that disrupt the regularity. This memory mechanism may allow for efficient and effective selection of information that needs to be memorized without awareness.
Discriminant Analysis Based on Naive Canonical Correlationsin High Dimension Low Sample Size
Mitsuru Tamatani
Faculty of Culture and Information Science, Doshisha University.
Abstract
In discriminant analysis for High Dimension Low Sample Size (HDLSS) setting itis not possible to define Fisher’s discriminant function, since the estimated covariancematrix is singular. HDLSS is characterized with large number of dimensions d and arelatively small number of sample size n, namely, we can express n ≪ d. Here, wepropose a new multi-class discriminant function based on naive canonical correlationsin HDLSS setting.
Consider K normal population with different means µℓ, and common covariancematrix Σ. Our proposal is to use the matrix
C = D−1/2E[(X − µ)Y T
] {E[Y Y T
]}−1/2,
where D = diagΣ, X is an observed vector to be classified and Y is the K-dimensionalclass label with ℓth entry 1 if X is from the ℓth class and 0 otherwise. Note that, if wereplace D by Σ, C plays an important role in canonical correlation analysis between Xand Y . The vector bi = D
−1/2pi (i = 1, . . . , K−1) is ith maximizer of the naive Fisher’sdiscriminant criterion, where pi is the eigenvector of the matrix CC
T belonging to theith largest eigenvalue. For two-class problems, the vector b1 becomes D
−1(µ1−µ2). Adiscriminant function based on b1 and called the naive Bayes has been proposed andinvestigated, see Bickel and Levina (2004), Fan and Fan (2008), and Tamatani, Kochand Naito (2012). We extend their results to a multi-class setting. In the generalisationfrom two-class to multi-class, the discriminant function g : Rd → {1, . . . , K} becomes
g (X) = argminℓ∈{1,...,K}
(X − µℓ)T B(BTΣB)−1BT (X − µℓ) , (1)
where B = [b1, . . . , bK−1] = D−1/2[p1, . . . ,pK−1]. Tamatani, Naito and Koch (2013)
derived an upper bound for the error probability in (1) in a HDLSS setting, that is,the dimension d and the number of sample size n satisfy n = o(d) as n, d → ∞. In thistalk, we discuss consistency properties of estimators of pi and bi in a HDLSS settingunder general distributional assumptions.
A Quantitative Comparative Analysis for Saikaku and Dansui’s Works
AYAKA UESAKA
Graduate school of Culture and Information Science ,Doshisha University
JSPS Research Fellow
ABSTRACT
In this article, we focus on Saikaku Ihara and Dansui Houjyou's writing s tyle.
Saikaku Ihara (c. 1642–93) is one of the most famous writers of the Genroku period
(1688 ‐ 1704) in Japan. Saikaku’s works are known for their significance for
developing Japanese novels today. However, it remained unclear which works were
really written by Saikaku. Meanwhile, the potential of quantitative analysis of textua l
data has dramatically advanced. Based on the new capabilities, we examined Saikaku
and Dansui's works using Welch's t -test and Principal Component Analysis. These
results revealed that Saikaku and Dansui's works differ grammatically. Furthermore, we
analyzed particles characteristically used in each writer by using Welch's t-test and
PCA. These results also indicate that Saikaku and Dansui's works are quite different.
I. INTRODUCTION AND PURPOSE OF THIS STUDY
Saikaku Ihara (c.1642‐93) is a haikai poet and fiction writer of the Genroku period
(1688‐1704) in Japan. After publishing the maiden works of Koushoku ichidai otoko
(The Life of an Amorous Man;1682) , he became the leading author of Ukiyozoushi¹.
Saikaku is considered to be a writer representing Genroku period along with Monzaemon
Chikamatsu (1653‐ 1724) and Basho Matsuo (1644‐ 1694). In the late eighteenth
century, there was a Saikaku revival, inspiring many modern Japanese writers. Saikaku’s
works are known for their significance in developing Japanese novels today (Emoto and
Taniwaki, 1996).
Identifying authors of novels in this era is difficult because typically they do not
have original signatures. It is said that Saikaku wrote twenty-four works in 10 years.
Among these twenty-four works, almost half of them do not have handwritten signatures
or signature seals. However, those achievements have not been fully verified, due to
some doubts in their authorship. His researchers have tried to identify his works by
investigating their history, content, format, and other features of the works but problems
continue to exist. It remains unclear which works were really written by Saikaku
especially his five posthumous works (Saikaku oki miyage, Saikaku oridome, Saikaku
zoku turezure, Yorozu no humihougu and Saikaku nagori no tomo ) that were edited and
published from 1693 to 1699 by his student, Dansui Houjyou (1663-1711). Therefore,
there are suspicions that Dansui may have modified Saikaku‘s work.
In this paper, we focus on Saikaku and Dansui’s works to find a difference of writing
style using a quantitative approach. Accordingly, we used a quantitative approach to
inspect Saikaku’s authorship problems because the potential of quantitative analysis of
textual data has dramatically advanced. That method has made it possible to provide new
Willingness to Sacrifice for the Environment: A Comparison of Environmental Consciousness in China, Japan and South Korea
Yanyan Chen11 and Yuejun Zheng2 1Graduate School of Culture and Information Science, Doshisha University,
2Faculty of Culture and Information Science, Doshisha University,
Abstract: This paper aims to compare and clarify people’s environmental consciousness in China, Japan and South Korea, from a perspective of willingness to sacrifice (WTS) for the environment. Analytical results derived from cross-national survey data indicate that the Chinese, especially Beijing citizens, tend to hold positive WTS, the Japanese tend to hold negative WTS, and South Koreans are located in the middle. Based on the revised norm-activation theory, environmental anxiety and environmental responsibility judgments are proposed in this study to explain the formation of WTS on a national level. In addition, the influence of demographic factors to the formation of WTS is also clarified and the attributive features of individuals who are inclined to sacrifice for the environment are clarified on a national level. Keywords: Sacrificial willingness, Cross-national comparison, Norm-activation theory, Categorical data analysis
1. Introducti
Self-interest is traditionally identified as a major source of environmental problems (Hardin &Baden, 1977; Mansbridge, 1990, quoted in De Young, 2000). The formation of altruistic or self-sacrificing motives is supposed to be particularly important in leading people to behave in an environmentally way. This paper focuses on people’s sacrificial willingness and tries to measure and analyze such sacrificial willingness from a somewhat comprehensive perspective which involves money, life and even policy contexts. Taking the work of Diet, Stern and Guagnano (1998) as a reference, we term the sacrificial willingness in this study as “willingness to sacrifice” (WTS) for the environment.
WTS, as the name implies, is a kind of personal sacrifice that will benefit other members in society, especially future generations. The model of self-interest theory supplies little explanation for this personal sacrificial willingness while the norm-activation theory proposed by Schwartz (1970, 1977) is assumed to be helpful in explaining the formation of WTS. Environmental anxiety which is taken as a measurement of AC, and environmental responsible judgments which is taken as a measurement of AR, are supposed to affect the formation of WTS.
1 Correspondence concerning this article should be addressed to: Yanyan Chen, E-mail: [email protected]
Verification of ghostwriters in Yasunari Kawabata’s novel:―who are standing behind the winner of the Nobel Prize for literature
HaoSun,Jongchan Lee†and Mingzhe, Jin‡
dAebbsattreadc tb—y lAilttehroauryg hs cthhoel airsss, uteh eo f qguheostsitownr ihtearss naost soyceita tbeede nw ritehs olYvaesdu dnuarei toKaa wlaabcakt ao
analysis perspective is provided in this study. Textual measurements such as character–symbolbigrams, tag bigrams, particle bigrams, phrase patterntss , paronvdi dae nweeiwg hetveidd einncteetgor aptreodv ec ltahsasitf icOattoiomne anlogomriitnhamto awraes unsoetd wirni totuern abtyt rYibaustuinoan rim Ketahwoadb. aTtha ea lroenseu.l
1 Introduction
Verification of ghostwriters in Yasunari Kawabata’s novels: who are standing behind the winner of the Nobel Prize for literature
HaoSun ,Jongchan Lee†and Mingzhe Jin‡
Abstract:
Although the issue of ghostwriters associated with Yasunari Kawabata has long been debated by literary scholars, the question has not yet been resolved due to a lack of evidence. The traditional method of investigating the issue is by using historical records for verification. Letters between Yasunari Kawabata and ghostwriters or the manuscripts of ghostwriters are recognized as compelling evidence. Providing an alternative to traditional methods, new evidence from a data analysis perspective is provided in this study. Textual measurements such as character–symbol bigrams, tag bigrams, particle bigrams, phrase patterns, and a weighted integrated classification algorithm are used in our attribution method. The results provide new evidence to prove that Otome no minato was not written by Yasunari Kawabata alone.
1 Introduction
Yasunari Kawabata (1899–1972) was a famous Japanese novelist who won the Nobel Prize for Literature in 1968. He is known to have suffered from anxiety and required sleeping pills. Tragically, he chose to end his life in 1972. It has long been argued that he made use of ghostwriters. Tsuneko Nakazato, one of Yasunari Kawabata’s disciples, is said to be one of the authors of Yasunari Kawabata’s novel for girls, Otome no minato. To date, two pieces of evidence have been used to show that Tsuneko Nakazato is also one of the authors. The first is a part of the Otome no minato manuscript (Chapter 7) found in 1989, which was written by Tsuneko Nakazato. The second is advice about how to write Otome no minato in a letter from Yasunari Kawabata to Tsuneko Nakazao. However, this novel is still part of Yasunari Kawabata’s collected writings although evidence suggests that Otome no minato ought to be a collaborative work.
Statistical and computational methods in Japanese authorship attribution have a long history. They were first used in the 1950s to study Biten Yasumoto’s authorship of the Genjimonogatari. Over the past decade, this field has dramatically developed, taking advantage of research in machine learning and natural language processing, and, in particular, the emergence of Japanese tokenizer/parser tools. Since then, several types of textual measurements and algorithms for Japanese authorship attribution have been proposed. Four effective textual measurements are proposed from different aspects of linguistic theory. These are character–symbol bigrams, tag bigrams, particle bigrams, and phrase patterns [1-3]. To date, the newest and most effective attribution method is the integrated classification algorithm [4].
In this study, to eliminate contingency, all the textual measurements mentioned above and a weighted integrated classification algorithm were used to verify the association of ghostwriters with Otome no minato.
2 Method
There were three necessary steps for verifying the actual author of Otome no minato. First, a corpus of the works of Yasunari Kawabata and Tsuneko Nakazato was created. Second,
Graduate School of Culture and Information Science, Doshisha University, 1-3 TataraMiyakodani, Kyotanabe-shi, Kyoto-fu 610-0394, Japan, E-mail: [email protected]. †Visiting scholar of Faculty of Culture and Information Science, Doshisha University, 610-0394, Japan. ‡Faculty of Culture and Information Science, Doshisha University, 610-0394, Japan.
Canonical covariance analysis for three-modethree-way data based on Tucker model
Jun Tsuchida* and Hiroshi Yadohisa**
*Graduate School of Culture and Information Science, Doshisha University**Department of Culture and Information Science, Doshisha University
1 Introduction
Three-mode three-way data is observed by the same set of objects and variables indifferent source or conditions, with such data often obtained as a set of multivariatematrices. For example, panel data often include values for the same objects and vari-ables at different times. These data are often organized as a three-way array. Whenwe analyze these data, we frequently apply dimension reduction techniques becausethe number of variables and conditions are large. There are numerous dimensionalreduction for three-mode three-way data. The Well-known dimension reductionmethods include Tucker model (Tucker, 1966) (Kroonenburg, 1983) and PARAFACmodel (Harshman, 1970). Furthermore, Vichi et al. (2007) proposed simultaneousclustering and dimension reduction methods for three-mode three-way data based onTucker model. Given only one dataset,these methods have been proven to be effec-tive. Given two datasets, we can apply these methods to each dataset individually,but we can not interpret the relationship between these datasets.
On the other hand, canonical covariance analysis has proposed to investigate therelationships between two variable sets. As one such dimension reduction method fortwo-mode two-way data, and canonical covariance analysis attempts to find a sub-space such that the covariance between variables is maximized. When datasets areconverted via Mahalanobis transformation, canonical covariance analysis is equiv-alent to canonical correlation analysis (Hotelling, 1936). Carrol (1968) proposedgeneralized canonical correlation analysis for a set of multivariate matrices in whichthe matrices is unfolded by condition. However, because this method was pro-posed for two-mode two-way data, it does not consider conditions when reducingthe dimension. Thus in this method, the same variable under different conditions isregarded as different variables which makes interpretation difficult.
In this paper, we therefore propose canonical correlation analysis for three-modethree-way data. Using our proposed method, dimensional reductions in both con-dition and variable are simultaneously applied on the basis of Tucker model. Inaddition, variables that have different conditions can be interpreted as the samevariable by expressing the difference in condition as a linear combination.
Sampling Design for CGSS Yongjin Jin1
1. Scholl of Statistics, Renmin University of China
Abstract: In my talk, I’ll give a brief introduction of sampling design for CGSS (China
General Social Survey). CGSS is an influential social survey in China, even in the world on
sociological study. In this talk, I’ll introduce the stratified sampling, sampling stage, sample
size and sample size allocation, and to discuss that how we deal with problems when we face
these challenges.
Keywords: Sampling stage, Sampling size, allocation of sample.
Robust Model-Free Feature Screening via Quantile
Correlation Xuejun Ma1, Jingxiao Zhang1
1. Center for Applied Statistics, School of Statistics, Renmin University of China
* Corresponding author: Email:[email protected]
Abstract: In this paper, we propose a new sure independence screening procedure based on
quantile correlation(QC-SIS). The method not only is robust against outliers,but also can
discover the nonlinear relationship between independent variables and dependent variable.
We establish the sure screening property under certain technical conditions.Simulation
studies are conducted to assess the performances of QC-SIS,sure independent screening(SIS),
sure independent ranking and screening(SIRS), robust rank correlation screening(RRCS),
distance correlation-sure independent screening(DC-SIS) and martingle difference
correlation-sure independence screening(MDC-SIS). Results have shown the effectiveness
and the flexibility of the proposed method. We also illustrate the QCSIS through an empirical
example.
Keywords: Quantile correlation, Ultrahigh-dimensionality, Sure screening, Robustness.
A Choice Model with a Diverging Choice Set for POI
Data Analysis Xiaoling Lu1, Julong Zhao2, Yu Chen3 and Hansheng Wang3
1. Center for Applied Statistics Research, Data Mining Center, School of Statistics, Renmin
University of China
2. School of Mathematics and Systems Science, Beihang University
3. Guanghua School of Management, Peking University
Abstract: A point of interest (or POI) is a geographical location, which might carry some
interest to the public. POI provides a convenient way to register people's locations through
mobile devices, which leads to POI data. POI data contains accurate location information and
is extremely valuable for location based services (LBS) and others. Accordingly, principled
statistical methods, which can be used for regression and/or prediction in this regards, are
needed. To partially fulfill this theoretical gap, we propose here a conditional logit approach
for POI choice analysis. The new model is a natural extension of the classical choice model
(McFadden, 1974, 1978) but with two key characteristics. First, POIs located far away from
the current position are less likely to be selected as the next POI choice. As a result, the
distance (or its appropriate transformation) between the current position and the next POI
candidate is an important predictor and should be always included in the proposed model.
Second, the classical choice model considers a finite choice set. In contrast, the new model
studies a diverging choice set, mainly due to the fact the total number of POI locations in
practice is typically huge. The diverging choice set makes the computation of the maximum
likelihood estimation (MLE) expensive. To alleviate computational cost, we further propose a
constraint maximum likelihood estimation (CMLE) method. Compared with MLE, CMLE
utilizes only those POIs located within a reasonable distance. This leads to significant
reduction in computation, at a reasonable cost of efficiency loss. To demonstrate the
method's finite sample performance, numerical studies based on both simulated and real
datasets are presented.
Keywords: Choice Model, Constrained Maximum Likelihood Estimation, Diverging Choice
Set, Location Based Service, Point of Interest.
Estimating optimal treatment regimes from a
classification perspective Baqun Zhang1,*, Anastasios A.Tsiatis2, Marie Davidian2, MinZhang3 and Eric Laber2
1. Department of Preventive Medicine , Northwestern University, Chicago,IL60611,USA
2. Department of Statistics, North Carolina State University, Raleigh, NC, 27695-8203,USA
3. Department of Biotatistics, University of Michigan, Ann Arbor, MI48109-2029,USA
* Email:[email protected]
Abstract: A treatment regime maps observed patient characteristics to a recommended
treatment. Recent technological advances have increased the quality, accessibility, and
volume of patient-level data; consequently, there is a growing need for powerful and flexible
estimators of an optimal treatment regime that can be used with either observational or
randomized clinical trial data. We propose a novel and general framework that transforms
the problem of estimating an optimal treatment regime into a classification problem wherein
the optimal classifier corresponds to the optimal treatment regime. We show that commonly
employed parametric and semi-parametric regression estimators, as well as recently
proposed robust estimators of an optimal treatment regime can be represented as special
cases within our framework. Furthermore, our approach allows any classification procedure
that can accommodate case weights to be used without modification to estimate an optimal
treatment regime. This introduces a wealth of new and powerful learning algorithms for use
in estimating treatment regimes. We illustrate our approach using data from a breast cancer
clinical trial. Copyright© 2012JohnWiley&Sons,Ltd.
Keywords: Classification, Doubly robust estimator, Inverse probability weighting,
Personalized medicine, Potential outcomes, Propensity score.
Adaptive Quantile Regression and Its Application
to VaR Maozai Tian1 and Ngai Hang Chan2
1. School of Statistics, Lanzhou University of Finance and Economics,Lanzhou,730101,Gansu
1,2. Center for Applied Statistics, School of Statistics, Renmin University of China,Beijing,100872,
China
2. Department of Statistics, Chinese University of Hong Kong, Shatin, Hong Kong
Abstract: An adaptive local smoothing method for nonparametric conditional quantile
regression models is considered in this paper. Theoretical properties of the procedure are
examined. The proposed method is fully adaptive in the sense that no prior information
about the structure of the model is assumed. The fully adaptive feature not only allows
varying bandwidths to accommodate jumps or instantaneous slope changes, but also allows
the algorithm to be spatially adaptive. Under general conditions, precise risk bounds for
homogeneous and heterogeneous cases of the underlying conditional quantile curves are
established. An automatic selection algorithm for locally adaptive bandwidths is also given,
which is applicable to higher dimensional cases. Simulation studies and data analysis
confirm that the proposed methodology works well.
Keywords: Adaptive Smoothing, Automatic Bandwidths, Conditional Quantile, Risk Bounds,
Robustness.
1
Academia, Salary or Family, Talent Mobility in and across China
Feng Zhen1 1. School of Statistics, Renmin University of China, Beijng, China – [email protected]
Abstract: Statistics shows that Chinese government spend quite a lot of fund to support talents from all over the world to work for the fast growing country. Some people argue that the fund is in a low efficiency. The government also realised the problem, especially some of the frequent mobility, and plan to build a survey system to understand the reason of talent moving and efficiency of the fund. The analysis of talents mobility helps standardize the system of human resources management, boost tap full potential of high-end talents, and reduce the costs of personnel administration. Based on the first roll survey of high-end talents mobility organized by the government in 2013, we studied the factors influencing decision-making policy of talents mobility and times of staff transfer among colleges and universities, or between academy and non-academy. The empirical results suggest that external factors such as family or personal affairs have the most powerful influence, while the pursuit of better academic capability and academic atmosphere or even higher salaries have less influence. The platform, development and atmosphere of academia have not been regarded as the priority during the allocation of human resources in current situation. Perceiving the attitudes of HR Office, as well as personal attitudes towards personnel mobility itself directly affect one’s decision-making in talents’ transfer.
Key words: Talents mobility, Survey, Decision-making, Zero-inflated negative binomial model.
mailto:[email protected]
Variable Selection in Semiparametric Models for the Strong
Hierarchical Longitudinal Data Xianbin Zeng1,2, Shuangge Ma1,3, Yichen Qin4, Yang Li1,2,5,*
1. School of Statistics, Renmin University of China, Beijing, P.R. China
2. Statistical Consulting Center, Renmin University of China, Beijing, P.R. China
3. Department of Biostatistics, Yale University, New Haven, U. S. A.
4. Department of Operations, Business Analytics and Information Systems, University of
Cincinnati, Cincinnati, U. S. A.
5. Center for Applied Statistics, Renmin University of China, Beijing, P.R. China
* Corresponding Authors: Yang Li, Associate Professor of School of Statistics, Renmin University
of China, [email protected].
Abstract: In this paper, we consider the variable selection problem in semiparametric additive
partially linear models for longitudinal data. Our goal is to identify relevant main effects and
corresponding interactions associated with the response variable. Meanwhile, we enforce the
strong hierarchical restriction on the model, that is, an interaction can be included in the model
only if both the associated main effects are included. Based on B-splines basis approximation for
the nonparametric components, we propose an iterative estimation procedure for the model by
penalizing the likelihood with a partial group MCP penalty, and use BIC to select the tuning
parameter. To further improve the estimation efficiency, we specify the working covariance matrix
by maximum likelihood estimation. Simulation studies indicate that the proposed method tends to
consistently select the true model and works efficiently in estimation and prediction with finite
samples, especially when the true model obeys the strong hierarchy. Finally, the China Stock
Market data are fitted with the proposed model to illustrate its effectiveness.
Keywords and phrases: Variable selection, Interaction, Semiparametric additive partially linear
model, Strong hierarchy, Longitudinal data.
Improved Semiparametric Estimation of Treatment Effect with Density
Ratio Model
Cunjie Lin , Wenhua Wei and Yong Zhou , 1. School of Statistics, Renmin University of China, Beijing, China2. Academy of Mathematics and Systems Science, Chinese Academy of Sciences,
Beijing, China3. School of Statistics and Management, Shanghai University of Finance and
Economics, Shanghai, China
Abstract: Treatment effect is an important index in comparing two-sample data in survival analysis, industry manufacture, clinical medicine and many other applications. In this study, we propose a unified semiparametric approach to estimate various indices of treatment effect under the density ratio model assumption, which directly connects two density functions by an exponential tilt. For each index of treatment effect, we construct two estimating functions using the auxiliary information extracted from the two-sample model and apply the generalized method of moments to improve the estimates. The estimating functions are allowed to be nonsmooth with respect to parameters, which makes it possible to evaluate the treatment effect from various perspectives simultaneously. We use the theory of empirical processes to prove the asymptotic properties of the proposed estimates and conduct some simulation studies to assess the finite sample properties. At last, we demonstrate the application of the proposed methods to the practical problems.
Keywords and phrases: Density ratio model; Empirical likelihood; Estimating equation; Generalized method of moments; Treatment e ect.
Tamatani_abstract2空白ページ空白ページ空白ページ
renda_abstraYongjin Jin_abstractJingxiao Zhang-abstractXiaoling Lv-abstractMaozai Tian-abstractBaqun Zhang-abstractFeng Zhen- abstractYang Li_abstractCunjie Lin-abstract