SELECTING PREDICTOR VARIABLES IN 2-GROUP CLASSIFICATION PROBLEMS

Citation
Jd. Morris et A. Meshbane, SELECTING PREDICTOR VARIABLES IN 2-GROUP CLASSIFICATION PROBLEMS, Educational and psychological measurement, 55(3), 1995, pp. 438-441
Citations number
11
Categorie Soggetti
Psychology, Educational","Psychologym Experimental","Mathematical, Methods, Social Sciences
ISSN journal
00131644
Volume
55
Issue
3
Year of publication
1995
Pages
438 - 441
Database
ISI
SICI code
0013-1644(1995)55:3<438:SPVI2C>2.0.ZU;2-4
Abstract
One approach to the problem of selecting predictor variables in the tw o-group classification situation is to compare the cross-validated cla ssification accuracies of all possible subsets. Because of the obvious time constraints of this approach for even a moderate number of predi ctors, a computer program was written to accomplish the task. The prog ram examines the cross-validated classification hit rates of all possi ble subsets of predictor variables and identifies the user-specified n umber of best subsets overall and for each subset size.