On the relative power of the U and t tests

Boneau (1962) concluded that there is little difference between the power of the Mann–Whitney U test and the t test when samples are drawn from non‐normal populations. He further concluded that, in the long run, the t statistic does hold a slight power advantage over the Mann–Whitney statistic in the non‐normal population situation. These conclusions are not supported by the asymptotic results obtained by Hodges & Lehmann (1956). In this study, computer simulation techniques were used to re‐examine a portion of Boneau's work. As a result of this re‐examination, it was demonstrated that Boneau erred in his conclusions and that the Mann–Whitney test may indeed show large power advantages over the t test.