Examining Source of Measurement Noninvariance in Multilevel Confirmatory Factor Analysis Using Multilevel Multiple Indicators, Multiple Causes (MIMIC) Modeling: A Simulation Study

• In Event: Poster Session 14
  In Poster Session: Advanced Issues in Multilevel Factor Analysis

Abstract

With the handful of research on cross-cultural measurement invariance (MI) using multilevel CFA (ML-CFA), researchers found that advanced levels of MI didn’t hold. To identify sources of noninvariance, researchers can utilize multilevel MIMIC (ML-MIMIC) model in which covariate effects are added to ML-CFA model. Through Monte Carlo simulation, this study aims to evaluate the performance of ML-MIMIC model to identify source of measurement noninvariance when biased items (DIF) are detected. Results showed that with medium DIF size, i.e., 0.5, and decent number of clusters, ML-MIMIC mostly detected DIF items. As number of clusters increased from 25 to 40 and 65, the difference in power due to ICCs became negligible. The cluster size seemed to exert no impact on the power.

Authors

• Abeer A. Alamri, University of South Florida

• Yan Wang, University of South Florida

• Eun Sook Kim, University of South Florida