Comparing the Performance of Multivariate Multilevel Modeling Using the Kenward-Roger Adjustment to Traditional Analyses

• In Event: 52.070 - Multilevel Modeling: Estimation, Model Selection, and Sample Size

Abstract

Multivariate multilevel modeling (MVMM) is receiving increased interest as an alternative approach to traditional multivariate analyses. The performance of MVMM, MANOVA, and a series of univariate tests was compared via a simulation study involving two groups and three dependent variables where the focus was on examining group mean differences. The MVMM tests were implemented with the Kenward-Roger adjustment. In addition, a new test for the multivariate null hypothesis under MVMM was proposed and its performance was evaluated. Across study conditions, MVMM provided greater power compared to traditional alternatives for both the test of the overall multivariate null hypothesis and tests for specific dependent variables. This procedure also provided accurate Type I error rates.

Authors

• Ryoungsun Park, Wayne State University

• Keenan A. Pituch, The University of Texas - Austin

• Jiseon Kim, University of Washington - Seattle

• Hyewon Chung, Chungnam National University

• Barbara G. Dodd, The University of Texas - Austin