学术动态

High-dimensional properties of model selection methods for estimating dimensionalities in multivariate linear model

报告人:藤越康祝

报告地点:数学与统计学院403室

报告时间: 2015年10月29日星期四14:00-15:00

 

主讲人简介:

藤越康祝(Fujikoshi Yasunori)教授是日本杰出的统计学家,他1970年从广岛大学获得博士学位,以后在神户大学任职到1978年,然后回到广岛大学数学系或应用数学系任教授直到2005年退休,后继续在广岛大学人荣誉退休(Emeritus)教授。迄今藤越教授发表重要杂志论文150余篇,专著10部。四十五年来,他先后担任多家重要国际杂志的副主编。由于他的杰出贡献,他自1980年期被选为日本统计学会和日本数学会理事,其中2003-2005年担任日本统计学会理事长。并多次获得日本国内奖励。他2004年被选为IMS Fellow。

 

报告摘要:

The problem of estimating dimensionalities in this talk involves the ones of estimating the reduced-rank in multivariate regression and the number of significant discriminant functions in discriminant analysis. As one of our main results we derive consistency properties of the estimation methods based on information criteria under a high-dimensional framework such that p/n tends to c<1, where n is the sample size and p is the number of response variables. For p>n, a modified method is proposed. We also give some properties based on simulation experiments.

专题网站Project site