Discrimination and Classification by Using the Median

Abstract

This study aimed to introduce an initial or acceptable solution to the discrepancy occurred between a model significance and low averages of the correct classification by discussing this problem in the two previous studies, on discrimination and classification, done by the researcher himself, using median instead of mean. The researcher suggested using a non-parametric technique for discrimination and classification by using median in certain cases. The study was divided into four areas: assumptions of discrimination analysis, solving the problems of classification averages in the researcher's two previous studies, a study proposal, and application. The study found that it was possible to get correct or acceptable averages of discrimination and classification by using the proposed model. That is, discrimination and classification by using median might be an initial, or acceptable at least, to identify the discrimination potentials of variables.