Evaluating Level-Specific Fit Indices in Multilevel Latent Growth Curve Model: A Monte Carlo Study

• In Event: Analyzing Longitudinal Data in Multilevel Modeling and Latent Growth Modeling

Abstract

Clustered longitudinal data (CLD) produced by longitudinal studies with cluster sampling designs characteristically feature a three-level hierarchical structure. Within the framework of multilevel structural equation modeling, the CLD can be analyzed with a two-level model, known as a multilevel latent growth curve model (MLGCM). Although MLGCMs are widely adopted to study growth trajectories at different levels, the question, how to evaluate the goodness-of-fit of within- and between-models in MLGCM?, has not been well addressed. A Monte Carlo study was conducted to assess the performance of level-specific fit indices in MLGCM in terms of their sensitivity to misspecifications occurring either in the marginal mean structure, in the between-covariance structure, or in the within-covariance structure. The findings and their implications are discussed.

Authors

• Hsien-Yuan Hsu, University of Texas Health Science Center at Houston

• Jr-Hung Lin, National Chiao Tung University

• Susan Troncoso Skidmore, Sam Houston State University

• Minjung Kim, The Ohio State University