PROCEEDINGS OF THE FIRST US/JAPAN CONFERENCE ON THE FRONTIERS OF STATISTICAL MODELING

 

VOLUME 2: MULTIVARIATE STATISTICAL

MODELING

 

EDITED BY

H. BOZDOGAN

 

KLUWER ACADEMIC PUBLISHERS 1994

DORDRECHT, THE NETHERLANDS

 

Ordering Address:

E-mail: Kluwer@world.std.com

 

® Designed by H. Bozdogan & Created by Hugh Bailey, 1991.

 
PREFACE TO VOLUME 2
ISBN 0-7923-2598-2 (Volume 2)

This volume contains the papers of of The US/Japan Conference on the Frontiers of Statistical Medeling: An Informational Approach which deal with multivariate statistical modeling.
Multivariate statistical analysis has come a long way and currently is in an evolutionary stage in the era of "Informational Modeling" techniques and high speed computation. The major problems with the conventional approach to multivariate modeling appears to be in obtaining the exact sampling distributions of the usual test procedures, the arbitrary assumptions made on the parameters, and the ever present "curse of dimensionality". The problem of finding the percentage points of the conventional test statistics in multivariate analysis has become rather difficult even with today's computational capabilities. New objective analysis techniques are needed in multivariate data analysis to overcome the existing difficulties with conventional techniques. Table look up is to become a thing of the past.
The papers in this volume discuss in detail some aspects of model-selection, and application and utilization of model selection via the informational approach to some difficult problems in multivariate modeling problems. These include how to determine the number of mixture clusters present in a data set, cluster analysis, modeling principal components, AIC-replacements of some multivariate tests of homogeneity, high dimensional covariance estimation, categorical data analysis by AIC, longitudinal data models with fixed and random effects, applications of multivariate time series analysis in biomedicine with feedback, simulation studies in classical ANOVA type models using the informational approach, latent trait models, and a review of applications in Psychometrics.
It is with great pleasure and gratitude that the editor and co-editors acknowledge the authors and the referees who contributed to this volume, and the chairs of this scientific core of the conference.
 
 
Hamparsum Bozdogan
The University of Tennessee
Knoxville, July 1993
 

CONTENTS OF VOLUME 2



 List of Contributors to Volume 2 vii
 Editor's General Preface ix
 Preface to Volume 2 xiii
Summary of Contributed Papers to Volume 2
1. Some Aspects of Model-Selection Criteria
S. L. Sclove
 37
2. Mixture-Model Cluster Analysis Using Model Selection Criteria and a New Informational Measure of Complexity
H. Bozdogan
 69
3. Information and Entropy in Cluster Analysis
H. H. Bock
 115
4. Information-Based Validity Functionals for Mixture Analysis
A. C. Cutler and M. P. Windham
 149
5. Unsupervised Classification with Stochastic Complexity
J. Rissanen and E. S. Ristad
 171
6. Modelling Principle Components with Structure
B. D. Flury and B. Neuenschwander
 183
7. AIC-Replacements for Some Multivariate Tests of Homogeneity with Applications in Multisample Clustering and Variable Selection
H. Bozdogan, S. L. Sclove, and A. K. Gupta
 199
8. High Dimensional Covariance Estimation: Avoiding The Curse of Dimensionality'
R. M. Pruzek
 233
9. Categorical Data Analysis by AIC
Y. Sakamoto
 255
10. Longitudinal Data Models with Fixed and Random Effects
R. H. Jones
 271
11. Multivariate Autoregressive Modeling for Analysis of Biomedical Systems with Feedback
T. Wada, T. Koyama, and M. Shigemori
 293
12.A Simulation Study of Information Theoretic Techniques and Classical Hypothesis Tests in One Factor ANOVA
E. P. Rosenblum
 319
13. Roles of Fisher Type Information in Latent Trait Models
F. Samejima
 347
14. A Review of Applications of AIC in Psychometrics
Y. Takane
 379