Sample size determination in the Mann–Whitney test

Abstract

This paper discusses the problem of determining the number of observations necessary to apply the nonparametric Mann-Whitney test. We describe the method given by Noether (1987) for determining a sample size which guarantees that the Mann-Whitney test at a given significance level α has a predetermined power 1-β. The presented theory is tested by calculating the empirical power in computer simulations. The paper also raises the issue of the method of rounding the determined sample size to an even number when the sample is divided into two equinumerous subsamples.