Generalizing Classification Tables to Large Populations

Abstract

In a mailing of an insurance offering to 303,000 credit customers in one of the large Midwestern states, the small number of responders necessitated a modification of the usual classification procedure. The researchers examined the technique of generalizing sample results in a classification study to large subpopulations of unequal sizes. The usual output from the discriminant analysis routine in SPSS was extended to handle the present statistical problems. Some of the advantages of the technique were discussed and a word of caution was noted.