A Regression Framework for Effect Size Assessments in Longitudinal Modeling of Group Differences

The use of growth modeling analysis (GMA)–particularly multilevel analysis and latent growth modeling—to test the significance of intervention effects has increased exponentially in prevention science, clinical psychology, and psychiatry over the past 15 years. Model-based effect sizes for differences in means between 2 independent groups in GMA can be expressed in the same metric (Cohen's d) commonly used in classical analysis and meta-analysis. This article first reviews conceptual issues regarding calculation of d for findings from GMA and then introduces an integrative framework for effect size assessments that subsumes GMA. The new approach uses the structure of the linear regression model, from which effect sizes for findings from diverse cross-sectional and longitudinal analyses can be calculated with familiar statistics, such as the regression coefficient, the standard deviation of the dependent measure, and study duration.