Updated on 2024/03/30

写真a

 
TAKAI,Keiji
 
Organization
Faculty of Business and Commerce Professor
Title
Professor
External link

Degree

  • 博士 ( 2009.3 )

Research Areas

  • Natural Science / Applied mathematics and statistics

Education

  • Osaka University   Graduate School, Division of Engineering Science

    - 2009

      More details

  • Osaka University   Graduate School, Division of Engineering Science

    - 2005

      More details

  • Osaka University   Faculty of Human Science

    - 2003

      More details

  • Osaka University   Graduate School, Division of Engineering Science

    2009

      More details

  • Osaka University   Graduate School, Division of Engineering Science

    2005

      More details

Professional Memberships

Papers

  • Model Selection with Missing Data Embedded in Missing-at-Random Data Reviewed

    Keiji Takai, Kenichi Hayashi

    Stats   6 ( 2 )   495 - 505   2023.4

     More details

    Publishing type:Research paper (scientific journal)   Publisher:MDPI AG  

    When models are built with missing data, an information criterion is needed to select the best model among the various candidates. Using a conventional information criterion for missing data may lead to the selection of the wrong model when data are not missing at random. Conventional information criteria implicitly assume that any subset of missing-at-random data is also missing at random, and thus the maximum likelihood estimator is assumed to be consistent; that is, it is assumed that the estimator will converge to the true value. However, this assumption may not be practical. In this paper, we develop an information criterion that works even for not-missing-at-random data, so long as the largest missing data set is missing at random. Simulations are performed to show the superiority of the proposed information criterion over conventional criteria.

    DOI: 10.3390/stats6020031

    researchmap

  • Incomplete-data Fisher scoring method with steplength adjustment Reviewed

    Keiji Takai

    Statistics and Computing   30 ( 4 )   871 - 886   2020.2

     More details

    Publishing type:Research paper (scientific journal)   Publisher:Springer Science and Business Media LLC  

    Abstract

    An incomplete-data Fisher scoring method is proposed for parameter estimation in models where data are missing and in latent-variable models that can be formulated as a missing data problem. The convergence properties of the proposed method and an accelerated variant of this method are provided. The main features of this method are its ability to accelerate the rate of convergence by adjusting the steplength, to provide a second derivative of the observed-data log-likelihood function using only the functions used in the proposed method, and the ability to avoid having to explicitly solve the first derivative of the object function. Four examples are presented to demonstrate how the proposed method converges compared with the EM algorithm and its variants. The computing time is also compared.

    DOI: 10.1007/s11222-020-09923-z

    researchmap

    Other Link: http://link.springer.com/article/10.1007/s11222-020-09923-z/fulltext.html

  • A Framework of ASP for Shopping Path Analysis Reviewed

    Katsutoshi Yada, Kei Miyazaki, Keiji Takai, Kohei Ichikawa

    Proceedings - 2017 4th Asia-Pacific World Congress on Computer Science and Engineering, APWC on CSE 2017   49 - 54   2018.10

     More details

    Language:English   Publishing type:Research paper (international conference proceedings)   Publisher:IEEE  

    In this paper, we propose an ASP system for shopping path analysis and a cloud based system with the necessary analysis service for marketing strategies that combines sales data with shopping path data. This paper explains a recommendation system based on position information or an application that calculates a basic marketing indicator using shopping path data by introducing a framework of ASP for shopping path analysis. Because the proposed ASP service is built on a cloud based system, the users can easily access the service through the web based interface and perform large-scale data processing using the computational resources of the cloud based system at low cost.

    DOI: 10.1109/APWConCSE.2017.00017

    Web of Science

    Scopus

    researchmap

  • On the use of the selection matrix in the maximum like- lihood estimation of normal distribution models with missing data Reviewed

    TAKAI,Keiji

    Communications in Statistics – Theory and Methods   47,14,3992-3407   2018

     More details

  • Global Distribution of Watches A network analysis of trade relations Reviewed

    Pierre-Yves Donze, Ken Ishibashi, Bo Wu, Yuta Kaneko, Kei Miyazaki, Keiji Takai

    2017 17TH IEEE INTERNATIONAL CONFERENCE ON DATA MINING WORKSHOPS (ICDMW 2017)   605 - 611   2017

     More details

    Language:English   Publishing type:Research paper (international conference proceedings)   Publisher:IEEE  

    The worldwide market for luxury and fashion goods is dominated today by a handful of multinational corporations (MNCs). The way MNCs access foreign markets and organize distribution, however, remains unclear. In this paper, based on an analysis of foreign trade statistics, we take the example of watches and provide a model to highlight the most important flows as well as regional hubs in this global distribution system. By using this matrix data about watch distribution as a network consisting of countries ( nodes) and trades (links), a network analysis is applied to extract hub nodes playing an important role. The network is visualized to represent the distribution system while focusing on heavily weighted links. As a result, the analysis has demonstrated that the flow of watches does not run always directly from the country of production to end consumers. Intermediaries play a key role, especially in regional markets like Asia, Europe and North America.

    DOI: 10.1109/ICDMW.2017.86

    Web of Science

    researchmap

  • Finite-sample analysis of impacts of unlabeled data and their labeling mechanisms in linear discriminant analysis Reviewed

    Kenichi Hayashi, Keiji Takai

    COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION   46 ( 1 )   184 - 203   2017

     More details

    Language:English   Publishing type:Research paper (scientific journal)   Publisher:TAYLOR & FRANCIS INC  

    It is widely believed that unlabeled data are promising for improving prediction accuracy in classification problems. Although theoretical studies about when/how unlabeled data are beneficial exist, an actual prediction improvement has not been sufficiently investigated for a finite sample in a systematic manner. We investigate the impact of unlabeled data in linear discriminant analysis and compare the error rates of the classifiers estimated with/without unlabeled data. Our focus is a labeling mechanism that characterizes the probabilistic structure of occurrence of labeled cases. Results imply that an extremely small proportion of unlabeled data has a large effect on the analysis results.

    DOI: 10.1080/03610918.2014.957847

    Web of Science

    researchmap

  • Shop area visit ratio, stay time, and sales outcomes: In-depth analysis based on RFID data Reviewed

    Zhen Li, Ken Ishibashi, Keiji Takai, Katsutoshi Yada

    2015 2nd Asia-Pacific World Congress on Computer Science and Engineering, APWC on CSE 2015   2016.5

     More details

    Language:English   Publishing type:Research paper (international conference proceedings)   Publisher:Institute of Electrical and Electronics Engineers Inc.  

    Prior research identifies that «visit» and «stay» relate positively to individual's purchase outcome. However, the mechanism that how do these two factors affect sales outcomes at an aggregated level is still unclear. This study posits that shop area visit ratio leads to sales outcomes through the effects on stay time, and the influence paths of shop area visit ratio and stay time on sales outcomes differ across products. To further clarify the mechanism, we propose a mediation model, to study how shop area visit ratio and stay time affect purchasers ratio and purchase volume. The data in our study are used from ratio frequency identification (RFID) technology, and matched with point-of-sale (POS) data. The major result suggests that for «selected products», the two sales outcomes (i.e., purchasers ratio and purchase volume) are mainly determined by stay time, while for «planned products», which depends more on shop area visit ratio than stay time. This finding can help retailers to visualize the process of customers' in-store shopping, and to better understand the importance of shop area visit ratio and stay time.

    DOI: 10.1109/APWCCSE.2015.7476231

    Scopus

    researchmap

  • Effects of unlabeled data on classification error in normal discriminant analysis Reviewed

    Keiji Takai, Kenichi Hayashi

    JOURNAL OF STATISTICAL PLANNING AND INFERENCE   147   66 - 83   2014.4

     More details

    Language:English   Publishing type:Research paper (scientific journal)   Publisher:ELSEVIER SCIENCE BV  

    Semi-supervised learning, i.e., the estimation of parameters based on both labeled and unlabeled data, is widely believed to be effective in constructing a boundary in classification problems. The present paper investigates whether this belief is true in the case of normal discrimination in terms of the classification error for normal and nonnormal data. For this investigation, we use the framework of missing-data analysis because data consisting of labeled and unlabeled individuals can be regarded as missing data. Based on this framework, we introduce two labeling mechanisms: feature-independent labeling and feature-dependent labeling. For each of these labeling mechanisms, we analytically derive the asymptotic relative efficiency based on the labeled data alone and based on both the labeled and unlabeled data. Numerical computations reveal that (i) under the feature-independent labeling mechanism, unlabeled data tend to contribute to the improvement of the classification error even for nonnormal data and (ii) under the feature-dependent labeling mechanism, unlabeled data from both normal and nonnormal distributions are helpful when the labeled data are informative, but unlabeled data can augment the classification error when the labeled data are not informative. Finally, we describe some future areas of research. (C) 2013 Elsevier B.V. All rights reserved.

    DOI: 10.1016/j.jspi.2013.11.004

    Web of Science

    researchmap

  • A framework for analysis of the effect of time on shopping behavior

    Keiji Takai, Katsutoshi Yada

    JOURNAL OF INTELLIGENT INFORMATION SYSTEMS   41 ( 1 )   91 - 107   2013.8

     More details

    Language:English   Publishing type:Research paper (scientific journal)   Publisher:SPRINGER  

    Due to technological developments, data about how many items a customer buys and how long the customer spends in a supermarket are available. A major problem with the data, however, is that there is no framework that considers the heterogeneity hidden in the data. In this article, we propose a framework that considers heterogeneity in the number of items a customer buys. The first step of our framework is based on the Poisson mixture regression model using a stationary time in the department where the items are sold as its independent variable. This model finds latent homogeneous groups of customers and gives the regression models within each group. It simultaneously classifies the customers into the homogeneous groups. In the second step of our framework, a method to investigate whether another factor (variable) influences the classification into homogeneous groups is presented. This proposed framework is applied to real data collected from the customers, and the effectiveness of the framework is shown. The managerial implications are drawn from the result of the analysis.

    DOI: 10.1007/s10844-012-0223-6

    Web of Science

    researchmap

  • Asymptotic inference with incomplete data Reviewed

    Keiji Takai, Yutaka Kano

    Communications in Statistics: Theory and Methods   2012

     More details

  • A framework for analysis of the effect of time on shopping behavior Reviewed

    高井啓二, 矢田勝俊

    Journal of Intelligent Information Systems   41   2012

     More details

  • Exploration of Dependencies among Sections in a Supermarket Using a Tree-Structured Undirected Graphical Model Reviewed

    Keiji Takai

    12TH IEEE INTERNATIONAL CONFERENCE ON DATA MINING WORKSHOPS (ICDMW 2012)   324 - 331   2012

     More details

    Language:English   Publishing type:Research paper (international conference proceedings)   Publisher:IEEE  

    In research of purchase behavior in a supermarket, it is important to understand dependencies among sections of the supermarket, with each section corresponding to a category of items. An undirected graphical model is a powerful tool for this purpose. A problem with the application of an undirected graphical model is that there are many variables and, thus, a lot of computation is needed. In this article, we first apply a tree-structured undirected graphical model to reduce the computational amount, and second, propose a method to impose a restriction, based on our needs, on the tree-structured undirected graphical model. The variables we use are the length of time spent in each section and the number of items bought from each section. We found that some of the sections have influence on the adjacent sections and that some of the other sections have no influence on the adjacent sections, but do have influence on the nonadjacent sections. We also found that the number of items and the length of stationary time in the sections that influence a large number of sections are negatively related to those same variables in the other sections. Based on this result, managerial implications are described. Finally, we summarize this article and discuss some problems in the application of graphical models.

    DOI: 10.1109/ICDMW.2012.105

    Web of Science

    researchmap

  • Analysis of NMAR missing data without specifying missing-data mechanisms in a linear latent variate model Reviewed

    Yutaka Kano, Keiji Takai

    JOURNAL OF MULTIVARIATE ANALYSIS   102 ( 9 )   1241 - 1255   2011.10

     More details

    Language:English   Publishing type:Research paper (scientific journal)   Publisher:ELSEVIER INC  

    It is natural to assume that a missing-data mechanism depends on latent variables in the analysis of incomplete data in latent variate modeling because latent variables are error-free and represent key notions investigated by applied researchers. Unfortunately, the missing-data mechanism is then not missing at random (NMAR). In this article, a new estimation method is proposed, which leads to consistent and asymptotically normal estimators for all parameters in a linear latent variate model, where the missing mechanism depends on the latent variables and no concrete functional form for the missing-data mechanism is used in estimation. The method to be proposed is a type of multi-sample analysis with or without mean structures, and hence, it is easy to implement. Complete-case analysis is shown to produce consistent estimators for some important parameters in the model. (C) 2011 Elsevier Inc. All rights reserved.

    DOI: 10.1016/j.jmva.2011.04.007

    Web of Science

    researchmap

  • Clockwise and anti-clockwise directions of customer orientation in a supermarket: Evidence from RFID data

    Marina Kholod, Keiji Takai, Katsutoshi Yada

    Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)   6883 ( 3 )   304 - 309   2011

     More details

    Language:English   Publishing type:Research paper (international conference proceedings)  

    Customer orientation is one of the important yet underresearched topics in the retailing management. In this paper we replicate and extend research by Groeppel-Klein and Bartmann (2008), analyzing the new type of data, namely RFID (Radio Frequency Identification) data, with the purpose to examine grocery shoppers' moving direction within the store and its influence on their buying behavior. As a result, we found out that attributes of shopping process, such as purchases and walking distance, vary significantly, depending on shoppers' clockwise or anti-clockwise moving direction. As retailers would benefit from knowledge about how the moving direction of customers relate to the buying behavior, managerial implications are proposed. © 2011 Springer-Verlag.

    DOI: 10.1007/978-3-642-23854-3_32

    Scopus

    researchmap

  • Finding Latent Groups of Customers via the Poisson Mixture Regression Model

    Keiji Takai, Katsutoshi Yada

    2011 IEEE INTERNATIONAL CONFERENCE ON SYSTEMS, MAN, AND CYBERNETICS (SMC)   3603 - 3608   2011

     More details

    Language:English   Publishing type:Research paper (international conference proceedings)   Publisher:IEEE  

    Due to developments in technology, movement data tracking a customer's movements in a supermarket in addition to conventional POS data are now available. A problem in analyzing such data is that an ordinary statistical model assuming customer homogeneity does not fit well to such data. In this article, we propose a framework for analyzing such data in a collection of supermarket departments. The framework is based on the mixture regression model assuming the customers' heterogeneity. By the model, we find the latent homogenous groups of the customers and explain the number of items by a stationary time based on the regression model in each latent group. The method of the mixture regression model is explained in addition to the estimation method. We found that a small number of the customers buy more items by going to the supermarket departments and are more sensitive to the stationary time, while a large number of the customers buy less and are less sensitive.

    Web of Science

    researchmap

  • Constrained EM algorithm with projection method Reviewed

    Keiji Takai

    Computational Statistics   1-14   2011

     More details

  • Relation between Stay-Time and Purchase Probability Based on RFID Data in a Japanese Supermarket Reviewed

    Keiji Takai, Katsutoshi Yada

    KNOWLEDGE-BASED AND INTELLIGENT INFORMATION AND ENGINEERING SYSTEMS, PT III   6278   254 - 263   2010

     More details

    Language:English   Publishing type:Research paper (international conference proceedings)   Publisher:SPRINGER-VERLAG BERLIN  

    Radio Frequency Identification (RFID) technology uses radio waves to track an object to which a small tag is attached. In a Japanese supermarket, we attach the RFID device to the cart and collect data on purchase behavior. In this article, we clarify the relation between purchase probability and the time customers spend in the store section by analyzing the RFID data with main use of descriptive methods. We clarify the way how the stay-time explains the purchase probability and characteristics of each store section. Based on the result, some implications for business are made as well.

    Web of Science

    researchmap

  • Simple Computation of Maximum Likelihood Estimates in Latent Class Model with Equality and Constant Constraints Reviewed

    Keiji Takai, Yutaka Kano

    COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION   38 ( 3 )   654 - 665   2009

     More details

    Language:English   Publishing type:Research paper (scientific journal)   Publisher:TAYLOR & FRANCIS INC  

    A simple computational method for estimation of parameters via a type of EM algorithm is proposed in restricted latent class analysis, where equality and constant constraints are considered. These constraints create difficulty in estimation. In order to simply and stably estimate parameters in restricted latent class analysis, a simple computational method using only first-order differentials is proposed, where the step-halving method is adopted. A simulation study shows that in almost all cases the new method gives parameter sequences monotonously increasing the Q-function in the EM algorithm. Analysis of real data is provided.

    DOI: 10.1080/03610910802604179

    Web of Science

    researchmap

  • Conditions for elimination of factor indeterminacy in time series factor analysis Reviewed

    高井 啓二

    New Trends in Psychometrics   443-452   2009

     More details

  • Test of independence in a 2 x 2 contingency table with nonignorable nonresponse via constrained EM algorithm Reviewed

    Keiji Takai, Yutaka Kano

    COMPUTATIONAL STATISTICS & DATA ANALYSIS   52 ( 12 )   5229 - 5241   2008.8

     More details

    Language:English   Publishing type:Research paper (scientific journal)   Publisher:ELSEVIER SCIENCE BV  

    Test of independence for 2 x 2 contingency tables with nonignorable nonresponses is discussed. Dependency assumption between two observed outcomes is required to achieve identification in many models with nonignorable nonresponses in the analysis of 2 x 2 contingency tables (e.g., [Ma, W.-Q., Geng, Z., Li, X.-T., 2003. Identification of nonresponse mechanisms for two-way contingency tables. Behaviormetrika 30, 125-144]). The assumption is, however, violated under the null hypothesis when implementing the test of independence. In this article, we introduce a new simple assumption to achieve identification. The assumption involves pre-specified parameters. EM algorithms for finding the MLE are numerically unstable when there are nonlinear constraints, which are created by models treating nonignorable nonresponses. In the analysis of contingency tables, estimated values often fall outside the admissible region. We propose a new EM type algorithm to stably calculate the constrained MLE, and apply it to make the test of independence for a real data set (crime data). We compare empirical performance among several testing procedures for independence. It turns out that the new EM type algorithm works well to calculate the MLE, and that the nonignorable model with the correctly specified parameters performs best while the conventional chi-square test of independence works fairly well. (c) 2008 Elsevier B.V. All rights reserved.

    DOI: 10.1016/j.csda.2008.04.029

    Web of Science

    researchmap

▼display all

Books

  • 欠測データの統計科学--医学と社会科学への応用--

    高井啓二, 星野崇宏, 野間久史( Role: Joint author)

    岩波書店  2016.4 

     More details

Research Projects

  • A Study of Cognitive Processes in Investor Decision Making Using Eye Tracking

    Grant number:22H00888  2022.4 - 2026.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (B)

      More details

    Grant amount:\16640000 ( Direct Cost: \12800000 、 Indirect Cost:\3840000 )

    researchmap

  • EMアルゴリズムに代わる欠測データを用いたパラメータ推定法の開発

    Grant number:18K11205  2018.4 - 2023.3

    日本学術振興会  科学研究費助成事業  基盤研究(C)

    高井 啓二

      More details

    Grant amount:\4160000 ( Direct Cost: \3200000 、 Indirect Cost:\960000 )

    本年度はいろいろな欠測データメカニズム下での情報量規準の開発と,完全にランダムな欠測の下での行列の一致推定の問題に取り組んだ.これまで開発されてきた情報量規準には応用上重大な問題点があった.それは情報量規準によって変数を選択するとき,欠測データメカニズムを一切考慮していないことであった.本来,欠測データメカニズムがランダムな欠測でない限り,情報量規準は適用できないはずである.しかし,従来の情報量規準を用いると,適用できないはずの欠測データメカニズムも選択してしまうことがあった.そこで,本年度は, 手に入る全てのデータを用いるときにはランダムな欠測であるという仮定をおいて,情報量規準を開発した.ここでおいた仮定は,変数が多ければ多いほど欠測を生じさせる変数がデータに含まれるという現実によく対応している.この仮定の下であれば,どのようなパラメータであっても一致推定可能になる.さらに,推定量の分散を小さくすることも可能となる.この情報量規準によって任意の変数のセットが選ばれても,パラメータは一致推定できるので,正確な情報量規準が計算できる.もう一つの研究では,完全にランダムな欠測の状況下での行列が,特にサンプルサイズが小さい場合に,非正定値になってしまう場合に対する対策法を開発した.このような問題が起きると,分散共分散行列が正定値であることを仮定している一般的な多変量解析法は一切使えなくなってしまう.対策としては,情報量をできるだけ減らさずに行列を正定値に変形することが重要になる.本年度は,非正定行列から最も近い正定値を固有値分解を用いて作成する方法を組み込んだアルゴリズムを作成した.

    researchmap

  • パスデータの融合による研究フロンティアの創出

    Grant number:16H02034  2016.4 - 2021.3

    日本学術振興会  科学研究費助成事業  基盤研究(A)

    矢田 勝俊, 高井 啓二, 宮崎 慧, 石橋 健, 李 振, 里村 卓也, 金子 雄太, 中原 孝信, 左 毅, 市川 昊平

      More details

    Grant amount:\38090000 ( Direct Cost: \29300000 、 Indirect Cost:\8790000 )

    多様なパスデータを用いた解析、モデル化を行うために、(1)消費者行動モデルの構築、(2)国際ワークショップの開催、(3)パスデータ融合の体系化に取り組み2019年度の計画通り研究を遂行することができた。
    (1)2018年度に実施した店舗実験で収集したパスデータをもとに、様々な消費者行動モデルを構築した。セルフコントロールが店内購買行為に与える時系列の影響についてモデル化に成功、また視線追跡を統合し、個人の異質性を前提とした確率モデルを提示することができた。
    (2)国際ワークショップの開催では、KES2019に併設する形で招待セッションを主催し、5本の発表を受入ながら、ブダペスト(ハンガリー)で行われた。顧客動線、視線追跡などの発表も含め、重要な意見の交換が行われた。そしてIEEEが主催する国際会議ICDMに併設する形で、Data mining for serviceという国際ワークショップを主催し、多くの発表を受入ながら、本提案の発表を行った。また、天野教授(ハーバード大学)やWedel教授(メリーランド大学)と議論しながら、体系化について貴重なアイデアを得ることができた。特にバースト検知手法を用いた混雑状況予測モデルについて、体系化が進み、論文化のめどをつけることができた。
    (3)体系化については、上述したような海外の研究者との議論を通して体系化のアイデアをまとめつつ、本提案チームの中で議論を深めることができた。ただし、3月に実施予定であった今期、最後の打合せが新型コロナウイルス感染症の感染拡大をうけて中止になった。

    researchmap

  • How serious is nonignorable missingness?

    Grant number:16K12402  2016.4 - 2018.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Challenging Exploratory Research

    Kano Yutaka, TAKAI Keiji, OTSU Tatsuo, MORIKAWA Kosuke, IMADA Miyuki, TAKAGI Yoshiharu, NAGASE Mario, Kim Jae-Kwang, Yuan Ke-Hai, Jamshidian Mortaza(Mori)

      More details

    Grant amount:\3380000 ( Direct Cost: \2600000 、 Indirect Cost:\780000 )

    Under NMAR missingness, the observed likelihood, without a missing-data mechanism, leads to a biased MLE. In this research, we developed a new methodology to express the bias of the MLE due to the missingness in closed form. Using the formula, we provided several mathematical conditions under which inclusion of auxiliary variables reduces or inflates the bias.
    The formula described above holds for any missing-data mechanism. This strong consequence can be proved because a shared-parameter model is taken for missingness.
    A final contribution of this research to be reported is to take a semi-parametric way to relax the strong condition required for the conventional missing-data analysis.

    researchmap

  • Study on the creation of a new ignobility condition and the investigation of an estimator distribution in the analysis of missing data

    Grant number:26730022  2014.4 - 2018.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Young Scientists (B)

    TAKAI Keiji

      More details

    Grant amount:\3770000 ( Direct Cost: \2900000 、 Indirect Cost:\870000 )

    In this project, I developed the theory for analysis of missing data and applied it to special data in the discriminant analysis. As theoretical study research, I derived conditional independences equivalent to MAR(missing at random) under monotonic and non-monotonic missing-data mechanisms. In addition, I constructed a method to overcome some difficulties in computation and estimation of the parameters of interest with missing data by using the selection matrix. It allows us to investigate properties of the estimator which are necessary for inference. As application research, I tackled a semi-supervised learning problem in discriminant analysis using the missing-data analysis theory. The semi-supervised learning is an estimation of the parameters from the partially observed data. I showed that the use of the missing-data analysis theory makes it possible to obtain the correct discriminant rule even with the such partially observed data.

    researchmap

  • New developments of missing data analysis: NMARness and APB

    Grant number:25540011  2013.4 - 2016.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Challenging Exploratory Research

    Yutaka Kano, IWASAKI Manabu, TAKAI Keiji, OTSU Tatsuo, HIROSE Kei, KAMATANI Kengo, KIKUCHI Kenichi, Sobel Michael E., Yuan Ke-Hai, Ricardo Silva, Mortaza Jamshidian, Aapo Hyvarinen

      More details

    Grant amount:\3640000 ( Direct Cost: \2800000 、 Indirect Cost:\840000 )

    This research project has been completed by the two research groups conducted by Professor Yutaka Kano and Professor Manabu Iwasaki. We have offered research colloquiums several times for each year to advance the research project. The aim of the research project is to re-structure the theory of missing data analysis and to apply them to some statistical models for the analysis with missing data. Results of the project include mathematically weakening the MAR condition, defining NMARness and Approximate population Bias (APB) and studying mathematical properties of the NMARness and APB. Applying these theoretical results, we studied effectiveness of introducing auxiliary variables in several statistical models for the analysis of missing data. One particular result is to derive mathematical conditions under which introducing surrogate endpoints can reduce the bias of the MLE for data with possibly missing data at the endpoint.

    researchmap

  • Development of Analysis Methods based on Generalized Methods of Moments for Nonignorable Missing Data

    Grant number:24700284  2012.4 - 2014.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Young Scientists (B)

    TAKAI Keiji

      More details

    Grant amount:\2860000 ( Direct Cost: \2200000 、 Indirect Cost:\660000 )

    As the first result of my study, I found that the maximum likelihood estimator (MLE) constructed from incomplete data has desirable properties such as consistency and asymptotic normality under the different conditions to the case in which completely observed data are available. The second result is application of the first result to discriminant analysis with partially labeled data. Since the partially labeled data can be regarded as missing data, the first result can also be applied to estimation of the parameters in such discriminant model when constructing a discriminant rule. It is found that all data available to us should be used when the observations used to construct the rule is completely randomly chosen, while there are times when not all observations with or without labels should be used for the data which are chosen depending on the value of the feature vector.

    researchmap

  • Experiment and Modeling of Consumer Behavior Analysis by Using Shopping Path Data

    Grant number:22243033  2010.4 - 2015.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (A)

    YADA Katsutoshi, KAWAKAMI Tomoko, NAKAHARA Takanobu, ICHIKAWA Kouhei, NISHIOKA Kenichi, TAKAI Keiji

      More details

    Grant amount:\32500000 ( Direct Cost: \25000000 、 Indirect Cost:\7500000 )

    The purpose of this research is to provide a model of consumer behavior which presents the relationship between consumer's in-store movement and purchasing behavior. We integrate customer's purchase history data with shopping path data which is consumer's in-store movement data in a retailer and then present various theoretical models such as customer's behavior model based on a string analysis, a time-series model of customer's movement and so on. Actually we implemented some marketing promotions in the real retailer based on our findings and achieved good sales performance. In conclusion we present a comprehensive framework of shopping path research based on these findings. We show significant implications for retail business and some theoretical contributions in marketing.

    researchmap

  • Statistical prediction, causation, incomplete data analysis and foundation of sciencee

    Grant number:22300096  2010.4 - 2014.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (B)

    KANO YUTAKA, DEGUCHI Yasuo, WASHIO Takashi, HAMAZAKI Toshimitsu, TAKAGI Yoshiji, SUGIMOTO Tomoyuki, TAKAI Keiji, NAITO Kanta, SHIMIZU Shohei, KATAYAMA Shota, YAMAMOTO Michio, SONG Xinyuan, JAMSHIDIAN Mortaza, HYVARINEN Aapo, YUAN Ke-hai

      More details

    Grant amount:\17810000 ( Direct Cost: \13700000 、 Indirect Cost:\4110000 )

    Analysis of incomplete data has been troublesome both theoretically and practically. In particular nonignorable missingness has been a serious issue in statistics. An alternative perspective of the theory of missing data analysis is to provide an insightful view of statistical causal inference. Some notable research outcomes include development of the analysis of doubly censored data, a new method of exploring causal structure for data with latent confounders via the LiNGAM approach, incomplete data analysis with a shared-parameter model and development of the EM algorithm with constraints.

    researchmap

▼display all