A new approach of information discrepancy to analysis of questionnaire data

Analysis of questionnaires plays an important role in many areas. In this article, a new approach based on-a measure of information discrepancy is proposed to investigate variable selection and discriminant analysis which are two basic problems in analysis of questionnaire data. In contrast to other methods, this approach can be used in cases with multiple groups of samples, i.e., it can deal with more complicated questionnaires involving multiple questions, multiple answers, and multiple groups. Moreover, the elementary applications show that the new approach of discriminant analysis is not only effective but also simple.