Comparing Decompose-First Methods for the 1-1-1 Mediation With a Moderator

• In Event: Methodological Challenges in Multilevel Modeling

Abstract

The purpose of this study is to compare the performance of methods for multilevel moderated mediation. The following approaches were compared as methods to test coefficients decomposed level by level: orthogonal partitioning approach with centering within cluster (OPC), random coefficient prediction (RCP), and latent moderated structural equations (LMS). Simulation studies were conducted, and manipulated conditions were analysis method, number of groups, group size, and intraclass correlation. The results showed that LMS and RCP can be recommended for within-level hypothesis and that LMS can be recommended for cross-level hypothesis. This study is meaningful in using the interpretable coefficients by applying the decomposed-first strategy in multilevel moderated mediation, and in extending the analytical model to include more specific research questions in MSEM.

Authors

• So Young Kim, Chonnam National University

• Sehee Hong, Korea University