Comparing Traditional Models and Random Intercept Models in Longitudinal Mediation

• In Event: Advances in Multilevel Modeling for Complex Data and Study Designs

Abstract

This simulation study compared traditional models (panel model and latent panel model) with random intercept models (random intercept panel model and trait-state-occasion model) in detecting longitudinal mediation effects. Four models were evaluated across varying sample sizes, intraclass correlations (ICC), and mediation effect sizes. Outcomes included convergence, power, type I error rate, and estimation accuracy. Results revealed that traditional models offered better convergence and greater statistical power, particularly when ICC was low. Alternatively, random intercept models, especially the trait-state-occasion model, yielded more accurate estimates when ICC was medium or high, and when larger sample sizes were available. The findings offer practical guidance on model selection between traditional and random-intercept models in longitudinal mediation analyses based on sample size and variance structure.

Authors

• Qiyue Zhang, Texas A&M University

• Paul R. Hernandez, Texas A&M University

• Oiman Kwok, Texas A&M University

• Wen Luo, Texas A&M University

• Karen E. Rambo-Hernandez, Texas A&M University