Sufficient Number of Clusters and Cluster Size for the Three-Level Random-Intercept Multivariate Multilevel Model

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

Abstract

The three-level multivariate multilevel model (MVMM) is a multivariate extension of the conventional two-level hierarchical linear model (HLM) and can be used to estimate and test the effects of predictors on a set of correlated continuous outcomes. The current simulation study examined the impact of number of clusters, cluster size, and other design characteristics on maximum likelihood (ML) parameter and standard error estimates, power, and type I error of three-level MVMMs with random intercepts and fixed slopes. The study also compared the MVMM results to those obtained from a series of HLMs. Results indicate that sample size requirements for the random-intercept MVMM do not differ from those for corresponding HLMs across the conditions investigated in this study.

Authors

• Wanchen Chang, Boise State University

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

• Susan Natasha Beretvas, The University of Texas - Austin