Theoretical considerations of model selection method based on a minimization of an information criterion

2015 Fiscal Year Final Research Report

• Project/Area Number: 25540012
• Research Category: Grant-in-Aid for Challenging Exploratory Research
• Allocation Type: Multi-year Fund
• Research Field: Statistical science
• Research Institution: Hiroshima University
• Principal Investigator: Yanagihara Hirokazu 広島大学, 理学(系)研究科(研究院), 准教授 (70342615)
• Co-Investigator(Renkei-kenkyūsha): FUJISAWA HIRONORI 統計数理研究所, 数理・推論研究系, 教授 (00301177) ; NINOMIYA YOSHIYUKI 九州大学, マス フォア インダストリ研究所, 准教授 (50343330)
• Project Period (FY): 2013-04-01 – 2016-03-31
• Keywords: モデル選択 / 情報量規準 / 多変量線形回帰モデル / 高次元データ

Outline of Final Research Achievements

In this research, we study model selection method based on a minimization of an information criterion. There are many information criteria. We evaluate theoretical properties of an information criterion used for model selection by a large sample asymptotic theory such that the sample size goes to infinity and a high-dimensional and large sample asymptotic theory such that the sample size and the dimension of a vector of response variables go to infinity simultaneously. From obtained results, we provide standards of judgment with regard to deciding an information criterion.

Free Research Field

統計科学