ON THE ANALYSIS OF MULTIPLE REGRESSION IN k CATEGORIES

Abstract

SUMMARY The general model, an information theoretic approach and solution to problems of test of hypotheses concerning sets of partial regression coefficients from k categories each involving p +1 variates, is presented and applied to certain data. The significance test is the analysis of variance ratio. It is shown that Carter's (1949) problem involving a ‘correlation effect’ among the ith members of each category reduces to a special case. The case of stochastic dependence among categories has been included by Kullback (1956) in his discussion of the multivariate linear hypothesis.