A Simulation Study of the Powers of Three Multiple Comparison Statistics

Summary A normal‐theory and two distribution‐free statistics used for multiple comparisons of homogeneity of location are compared on simulated data generated from six distributions. The normal‐theory statistic is found to be fairly robust to departures from the assumption of normally distributed data of the types considered. The Steel‐Dwass statistic is generally more powerful than a Kruskal‐Wallis range statistic.