Comparing Methods for Multilevel Moderated Mediation: A Decomposed-first Strategy

Abstract

The purpose of this study is to propose a decomposed-first strategy for multilevel moderated mediation and to compare the performance of three moderated mediation approaches in multilevel structural equation modeling. The following approaches were compared in simulations to test coefficients that were decomposed level by level: orthogonal partitioning with centering within cluster, random coefficient prediction, and latent moderated structural equations. The manipulated conditions for the simulation analysis were the analysis method, the number of groups, group size, and intraclass correlation. The results showed that, for samples consisting of a large number of groups, a large average group size and a large intraclass correlation, LMS had the strongest performance. This study is meaningful in that it produces interpretable coefficients by applying a decomposed-first strategy in multilevel moderated mediation and extends a basic moderated mediation model to include more specific research questions in multilevel structural equation modeling.

Keywords:

• Decomposed-first strategy
• latent moderated structural equations
• multilevel moderated mediation
• orthogonal partitioning with centering within cluster
• random coefficient prediction

Additional information

Funding

This work was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea [NRF-2019S1A5B5A02038811].