Distributions of Sums of Squares of Rank Differences for Small Numbers of Individuals

Abstract

In a recent article,2 reporting the results of research under a grant-in-aid from the Carnegie Corporation of New York, Hotelling and Pabst have given a comprehensive treatment of the theory and application of rank correlation and have contributed significantly to existing knowledge on the subject. It is not the purpose of this note to evaluate their contribution but to attempt the solution of a problem they suggest. In ?33 they have given the well-known formula for rank correlation, r' = 1 -