The 286th meeting of Hiroshima Statistics Study Group

Day & Time
31st October 2014, 15:00~
RERF Auditorium
Yasunori Fujikoshi, Ph.D.
Emeritus Professor of Hiroshima University,
Advisor of the Hiroshima University Statistical Science Research Core

Presentation Title
Estimation methods of the rank of mean parameter matrix in multivariate linear model and their high-dimensional consistency properties
This talk is concerned with the problem of estimating the rank of a mean parameter matrix in multivariate linear model. First we note that the problem involves some important problems as a special case. For example, it involves the problem of estimating the number of meaningful discriminant functions in discriminant analysis. Further, it is related to reduced-rank regression in multivariate linear regression analysis. As one of our main results we show consistency properties of some estimation criteria based on model selection method under a high-dimensional framework. We also review some other methods in high-dimensional case.