|
Abstract:
|
Statistical methods based on ranks are known to provide robust alternatives to the corresponding classical methods based on the assumption of normality of the data. Since under very general assumptions for an AVOVA model, linear functions of ranks are asymptotically normally distributed, it is recommended in the literature that the classical methods of multiple comparisons be directly extended to analogous functions of rank statistics. In this study, we investigate the veracity of this type of extension for some well-known multiple comparison methods in ANOVA models. Special focus will be given to the robustness of validity of rank-based multiple comparison methods for some small, moderate, and large designs.
|
