LOADINGS AND CORRELATIONS IN THE INTERPRETATION OF PRINCIPAL COMPONENTS

被引：232

作者：

CADIMA, J

JOLLIFFE, IT

机构：

[1] UNIV ABERDEEN, DEPT MATH SCI, ABERDEEN AB9 2TY, SCOTLAND

[2] INST SUPER AGRON, DEPT MATEMAT, P-1399 LISBON, PORTUGAL

来源：

JOURNAL OF APPLIED STATISTICS | 1995年 / 22卷 / 02期

关键词：

D O I：

10.1080/757584614

中图分类号：

O21 [概率论与数理统计]; C8 [统计学];

学科分类号：

020208 ; 070103 ; 0714 ;

摘要：

Principal components (PC) are often interpreted by looking at the loadings for each variable. Variables with a small-magnitude loading are ignored and the given PC is then approximated by the linear combination involving only the remaining variables. It is argued that this procedure is potentially misleading in various respects. Examples are given and alternatives are suggested.

引用

页码：203 / 214

页数：12

共 17 条

[1]

ARNOLD GM, 1993, J R STAT SOC C-APPL, V42, P381

[2]

Cailliez F, 1976, INTRO ANAL DONNEES

[3]

Chatfield C., 1980, INTRO MULTIVARIATE A

[4]

Draper N.R, 1998, APPL REGRESSION ANAL, V326, DOI 10.1002/9781118625590

[5]

ESCOUFIER Y, 1987, DEV NUMERICAL ECOLOG

[6]

Fenelon, 1982, TRAITEMENT DONNEES S

[7]

GOWER JC, 1966, BIOMETRIKA, V53, P325, DOI 10.2307/2333639

[8] Analysis of a complex of statistical variables into principal components [J].

Hotelling, H .

JOURNAL OF EDUCATIONAL PSYCHOLOGY, 1933, 24 :417-441

[9]

JACKSON JE, 1991, USERS GUIDE PRINCIPL

[10]

Jeffers J.N.R., 1967, APPLIED STATISTICS, V16, P225

← 1 2 →