This paper contains both a refinement and an extension of a distribution-free method of discriminate analysis proposed by Kendall and Stuart (1966). The Kendall and Stuart method is based on the assumption that, once a variable has been employed in the analysis, all of the discriminatory power or information of that variable has been exhausted. This assumption is shown to be not necessarily true,

... [Show full abstract] and a procedure is presented which enables one to utilize such information. Further, the Kendall and Stuart method is restricted to marginal frequency, distributions of possible discriminatory variables. It is shown that further discrimination may be possible by extending the examination to joint frequency distributions.