Canonical Analysis: A Factor-Analytic Method of Comparing Finite Discrete Distribution Functions

Abstract

The methods of factor analysis are adapted to the comparison of finite discrete distribution functions. Each element of a finite class of distribution functions is presented in a canonical form in which components differentiating and uniting the functions are easily distinguished and are readily calculated using factor-analysis algorithms. This technique, which we call canonical analysis, has uses in heterogeneity and goodness-of-fit analysis and in the classification of distribution functions.