A Comparison of Multilevel Mediation Modeling Methods

• In Event: Longitudinal Data, Model Evaluation, and Multilevel Models

Abstract

Mediation in multi-level data can be examined using conflated multilevel modeling (CMM), unconflated multilevel modeling (UMM), or multilevel structural equation modeling (MSEM). A Monte Carlo study was performed to compare the three methods on bias, type I error, and power in a 1-1-1 model with random slopes. The three methods showed no differences regarding the within indirect effect. When the between indirect effect was not directly simulated by the data generation model, UMM had an inflated type I error rate and UMM and MSEM exhibited positively biased estimates. When the between indirect effect was directly simulated, UMM and MSEM were both unbiased with UMM presenting higher power.

Authors

• Christina K Zigler, University of Pittsburgh

• Feifei Ye, University of Pittsburgh