A Monte Carlo Study of Analysis of Variance and Competing Rank Tests for Scheffé's Mixed Model

Abstract

Monte Carlo estimates are obtained of the small sample power of the classical F test and three competing rank tests for detecting two forms of treatment shift alternatives. These results were computed under Scheffé's mixed model for the case of one observation per cell using several variance-covariance matrices. Some small sample null distributions, empirical efficiencies, and robustness computations are also given.