Post hoc tests in factorial ANOVA: Why did SPSS remove the options for heterogeneous variances?

• In Event: Post Hoc Post Haste: Discussions on follow-ups to Analysis of Variance

Abstract

When there is a statistically significant main effect in a factorial ANOVA, researchers often wish to explore the data to determine where differences might exist. When the assumption of homogeneous variances is met with no apparent interaction, most of the same post hoc tests from one-way ANOVA are available to researchers. However, researchers don't have many options available to them for main effects post hoc tests with heterogeneous variances. For example, sometime after SPSS 18, these options were made unavailable in SPSS dialog boxes. The purpose of this paper is to investigate options for post hoc tests of statistically significant main effects in this heterogeneous variance case. Specifically, the Games-Howell test will be investigated along with another common post hoc approach to perform one-way ANOVA post hoc tests of the main effects. A series of Monte Carlo analyses will be conducted to investigate Type I error and power for these approaches under various conditions.

Authors

• Michael Senteney, Ohio University

• YI ZHOU, Ohio University

• Gordon Brooks, Ohio University