Variable selection for discrimination among several populations

被引:1
作者
Chatterjee, SK
Samanta, SK
机构
[1] UNIV CALCUTTA, DEPT STAT, CALCUTTA, W BENGAL, INDIA
[2] RK MISSION RESIDENTIAL COLL, NARENDRAPUR, W BENGAL, INDIA
关键词
noncentrality matrix; multiple discriminant analysis;
D O I
10.1080/03610929708832066
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
In this paper we develop a simple procedure for selecting a subset of variables for discriminating between several multivariate normal populations with the same known dispersion matrix. The procedure is based on consideration of the relative loss in trace of 'noncentrality matrix', resulting from the omission of some of the variables. It is so designed that the ultimate subset selected meets the requirement stipulated at a prescribed confidence level. The proposed procedure is a generalisation to the multipopulation case of one of the two-population procedures described in Chatterjee and Samanta (1992).
引用
收藏
页码:2565 / 2582
页数:18
相关论文
共 7 条
[1]  
CHATTERJEE SK, 1992, SANKHYA B, V54
[2]  
FUJIKOSHI Y, 1983, HIRO MATH J, V13
[3]  
GRIZZLE JE, 1983, CONTRIBUTION STATIST, P195
[4]  
Kshirsagar A. M., 1972, MULTIVARIATE ANAL
[5]   ON USE OF DISCRIMINANT FUNCTIONS IN TAXONOMY [J].
LUBISCHEW, AA .
BIOMETRICS, 1962, 18 (04) :455-&
[6]   SIMULTANEOUS PROCEDURES FOR VARIABLE SELECTION IN MULTIPLE DISCRIMINANT-ANALYSIS [J].
MCKAY, RJ .
BIOMETRIKA, 1977, 64 (02) :283-290
[7]  
RAO CR, 1973, LINEAR STATISTICAL I