Errors of misclassification associated with gamma distribution

Abstract

Errors of misclassification and their probabilities are studied for classification problems associated with two classes of univariate gamma distribution. The effects of applying the normal classificatory rule to gamma populations are also studied and assessed by comparing probabilities (optimum and conditional) based on the linear discriminant function (LDF) for normality with those based on the likelihood ratio rule (LR) for gamma populations. To determine the form in the onev-ariable case, formulas for the exact distribution and density functions of the actual error rates are presented. Both theoretical and empirical results are considered.