SELECTION OF COMPONENTS IN PRINCIPAL COMPONENT ANALYSIS - A COMPARISON OF METHODS

被引：101

作者：

FERRE, L ^{[1
]}

机构：

[1] UNIV TOULOUSE 3, STAT & PROBABIL LAB, CNRS, URA 745, F-31062 TOULOUSE, FRANCE

来源：

COMPUTATIONAL STATISTICS & DATA ANALYSIS | 1995年 / 19卷 / 06期

关键词：

PRINCIPAL COMPONENT ANALYSIS; FIXED EFFECT MODEL; CHOICE OF DIMENSIONALITY;

D O I：

10.1016/0167-9473(94)00020-J

中图分类号：

TP39 [计算机的应用];

学科分类号：

081203 ; 0835 ;

摘要：

The problem of the choice of the relevant components in principal component analysis is presented as a model selection problem. In this context, we present the numerous methods most often used to determine the number of relevant components and we try to show why unfortunately most of them fail. Then these methods are compared on simulated data to study their behaviour.

引用

页码：669 / 682

页数：14