Improved Robust Test Statistic Based on Trimmed Means and Hall's Transformation for Two-way ANOVA Models Under Non-normality

For the two-way fixed effects ANOVA, under assumption violations, the present study employs trimmed means and Hall's transformation to correct asymmetry, and an approximate test, such as the Alexander-Govern or Welch-James test, to correct heterogeneity. The unweighted as well as weighted means analyses of omnibus effects in unbalanced designs were considered. A simulated data set was presented and computer simulations were performed to investigate the small-sample properties of the methods. The simulation results show that the proposed technique is valid and powerful compared with the conventional methods.