Detailed Classification Problem

Abstract

Summary

The classification problem is concerned with allocation of new observations on several variables to two or more distinct categories known to exist a priori, on the basis of some pre-solved cases. Most of the existing solutions to the problem lack detailed inference with respect to the outcomes of the concerned random experiment. An alternative procedure is therefore suggested from frequentist viewpoint, involving relative credentials placed on the different candidate groups in terms of betting odds. The underlying theory was introduced in Chatterjee and Chattopadhyay(1993), the present study being an app lication towards the classification problem. Performance of the proposed ruie has been studied for some real-life examples through discriminant analysis.