A Monte Carlo Investigation of the Effectiveness of Backward Elimination Analysis as a Multivariate Analysis of Variance (MANOVA) Post Hoc Procedure

• Activity Number: 260
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Education
• Author(s): Chittanun Sitthisan*+
• Companies: University of Northern Colorado
• Keywords: Backward Elimination Analysis ; Multivariate Analysis of Variance (MANOVA) ; Post Hoc Procedure.

Abstract

Researchers have used univariate F-tests, descriptive discriminant analysis (DDA), and stepwise discriminant analysis (SWDA) as a post hoc procedure to a significant MANOVA. These studies have been shown to be inappropriate due to lack of power for some conditions through simulation studies. However, the backward elimination (BE) method could represent an improvement over SWDA in terms of detecting good subset equations. The purpose of this paper was to investigate the effectiveness of BE as a follow-up procedure to a significant MANOVA. Monte Carlo simulation was used to assess the effectiveness of backward elimination (BE) as a post hoc procedure to a significant MANOVA. Two conditions were examined using two different numbers of dependent variables (2, 5) and three levels of sample sizes (50, 150, 250), while controlling for effect size, correlation structures, and significance levels. The method appears to be a powerful procedure when the number of dependent variables is small and sample sizes are large. Finally, using BE with various conditions in future studies may enable researchers to further explore which conditions work best for BE as a follow-up procedure to MANOVA.