Abstract:
|
Nonparametric multiple comparisons allow researchers to compare multiple samples with virtually no distributional assumption on the data using the idea of relative effects. However, from the practical point of view, the raw estimated relative effect difference or its standardized version may be difficult to interpret, possibly making nonparametric multiple comparisons less attractive to practitioners. Thus, we suggest a modified approach that can accommodate various effect size measurements for comparing relative effects. In particular, we present modified test statistics and their finite-sample approximations. Moreover, we examine an application of the modified approach to a recent neuropsychological study.
|