A Simulation Study on Tests of Hypotheses for Fixed Effects in Mixed Models for One-Way ANOVA under the Violation of the Equal Variances Assumption of the Treatment Groups with and without Missing Data

This article considers the analysis of experiment of onewaycompletely randomized design (one-way ANOVA) that isfrequently used in every discipline. We investigate a common problemthat is data collected in practice usually violate parametricassumptions to some degree. We concentrated our attention onANOVA when the equal variances assumption for the treatmentgroups is violated. We investigate the performance of a general linearfixed effects model approach (GLM procedure of the SAS System) inanalyzing one-way ANOVA under the violation of only one assumptionthat is heterogeneous variances. Also, we investigate the performanceof a general linear mixed effects model approach (MIXEDprocedure of the SAS System) in analyzing one-way ANOVA underthe violation of only one assumption that is heterogeneous variancesas alternative to GLM procedure of the SAS System. The main resultof our article is that the general linear mixed effects model approachcan be recommended to be used in case of the suspicion of the violationof the equal variances assumption specially in case of unbalanceddata where the general linear fixed effects model approach showedserious departures upward from the nominal leve