Effect Size Measures for Longitudinal Growth Model

• In Event: Effect Sizes, Multilevel Models, and Causal inference

Abstract

R square (R2) is a popular approach to computing effect size measures. When applied to multilevel models (MLMs), the formulation is less straightforward, which resulted in various R2 measures developed for MLM. Recently, Rights and Sterba (2018) developed an integrative framework that allows researchers to choose effect size measures for key substantive questions in MLMs. However, the currently available R2 measures have limitations to be used for longitudinal growth models (LGMs). To overcome the limitations, the current study presents R2 formulation (a) that does not assume cluster-mean centered level-1 predictors, and (b) that provides partitioning of explained variance for any subset of fixed or random effects. The proposed R2 measures are illustrated using empirical data.

Authors

• Zonggui Li, Boston College

• Ehri Ryu, Boston College