A COMPARISON BETWEEN MAXIMUM LIKELIHOOD RULE AND LOGISTIC DISCRIMINANT ANALYSIS IN THE CLASSIFICATION OF MIXTURE OF DISCRETE AND CONTINOUS VARIABLES

Abstract

An optimal measure of performance is the one that lead to maximization of average error rate or probability of misclassification. This paper aimed to compare between the maximum likelihood rule and logistic discriminant analysis in the classification of mixture of discrete and continuous variables. The efficiency of the methods was tested using simulated and real dataset. The result obtained showed that the maximum likelihood rule performed better than the logistic discriminant analyses, in maximizing the average error rate in both experiment conducted.

Keyword: Maximum likelihood rule, Logistic discriminants, error rate, Likelihood ratio, Discriminant analysis.