“Be Congruent, Be Authentic, Be Your True Self”: Congruency of Robust MCPs (Multiple Comparison Procedures) With Robust ANOVA

• In Event: Machine Learning Methods in Educational Research

Abstract

Researchers are familiar with multiple comparison procedures (MCPs), like Tukey, Games-Howell, and Scheffé, used to explore pairwise group mean comparisons following a significant one-way ANOVA. Most have learned that some MCPs maintain Type I error rates better than others and some are more robust with violations of assumptions, while others may provide more power. Scheffé MCP lacks power but, uniquely, provides perfect congruence with Fisher’s one-way ANOVA (i.e., guarantees a significant comparison following a significant ANOVA). This suggests our research question: How congruent are robust MCPs (e.g., Games-Howell and the Brown-Forsythe adaptation of Scheffé) with robust one-way ANOVA (e.g., Welch), especially with violation of assumptions? Our purpose was to run Monte Carlo simulations to find congruence rates for robust MCPs.

Authors

• Gordon P. Brooks, Ohio University

• Nina Adjanin, Northwest Missouri State University

• Francis Ankomah, Ohio University

• Cletus Owusu Yeboah, Ohio University

• Emmanuel Awunkori, Ohio University