Content area
Abstract
Years ago, most researchers believed that parametric statistics were generally more powerful than their nonparametric counterparts when the assumptions of normality and homogeneity of variance were tenable. [...]despite any loss of power that may be experienced by the nonparametric Wilcoxon rank sum test, that test acquires a relative power advantage in outlierprone population distributions (Rasmussen, 1986; Zimmerman, 1994a, 1994b, 1996; Zimmerman & Zumbo, 1990). [...]those Monte Carlo simulations did not take into consideration (a) the situation in which the researcher uses a two-tailed decision rule or (b) the situation in which the researcher hypothesizes a one-tailed decision rule that is in the wrong direction. [...]when sample sizes were unequal the advantages of the nonparametric procedure in power, Type III error rates, and proportion of incorrect rejections were even more pronounced.