Comment on the Relative Usefulness of the Two Independent Means T-Test and Wilcoxon’s Rank-Sum Test for the Analysis of Educational Data
R. Clifford Blair · Florida Educational Research Association · 1979
Researchers and evaluators in education are often interested in assessing the effectiveness of some treatment through a test of the hypothesis f(x) = f(x-8), where f(x) is the functional form of the population(s) about which inferences are to be made and 8 is some constant that represents the treatment effect.Among educational researchers, by far the most common statlstic employed for this test is the two independent means t-test.This latter statement is true even though no specific or limiting conditions• were placed upon f(x).It will be contended here that this practice should be reexamined and that Wilcoxon's rank-sum (or the equivalent Mann-Whitney U) test should be substituted for the t-test in most of the situations where the latter statistic is routinely employed.It will be further contended that the present state of affairs was brought about by the exaggeration of facts and, more importantly, by an unclear understanding of the involved issues.