On Discrimination and Allocation with Continuous and Dichotomous Variables

Abstract

In discriminant analysis involving continuous and categorical variables, the simplest and conventional procedure is to assign an arbitrary numerical score to each possible state of the categorical variables and proceed as if all variables are continuous. A discrimination procedure is suggested for use in a situation where the discriminating variables are mixtures of more than one Continuous variable and one Dichotomous variable. The performance of the suggested procedure is compared alongside that of the conventional Fisher’s Linear Discriminant and Logistic Discrimination procedures based on their error rates. The suggested procedure performed better when compared with the other procedures. Hence, the suggested procedure will be applicable for such situation.

Keywords: Dichotomous, Continuous, Discriminant Analysis, Error Rate