A monte carlo study of rank tests for repeated measures designs

Abstract

Koch (1969) proposed a rank test to detect fixed treatment effects in complete repeated measures designs where the observations within blocks are assumed to be equally correlated. In deriving the limiting distribution of the rank transformation statistic for this design, Kepner and Robinson also obtained the limiting distribution of Koch's statistic under substantially less restrictive conditions than those required by Koch. The purpose of this paper is to summarize and disseminate theresults of a Monte Carlo study comparing the intermediate sample size performance characteristics of these two statistics, which until now has not been investigated, with two standard competitors,the analysis of variance (ANOVA) F test and Friedman's test

Keywords:

• Koch's statistic
• rank transformation statistic
• Friedman's test
• repeated measures
• within block correlation