Model Misspecification and Assumption Violations With the Linear Mixed Model: A Meta-Analysis

• In Event: Methodological Considerations in Modeling Latent Growth

Abstract

This meta analysis attempts to synthesize the monte carlo literature under a longitudinal framework. The empirical type I error rate will serve as the effect size and monte carlo simulation conditions will be coded to serve as potential moderator variables. The type I error rate for the fixed effects will be explored, but not for the random effects. Effect sizes were coded from 10 studies, resulting in a total of 3,776 effect sizes. Moderator variables related to the empirical type I error rate included the fitted serial correlation structure, missing a random effect, and the interaction between the fixed effect parameter and missing a random effect

Author

• Brandon LeBeau, University of Arkansas