A Monte Carlo Simulation Study: The Influence of Missing Data on Full Information Maximum Likelihood Estimation for Multilevel Structural Equation Modeling

• In Event: AERA Poster Session 6
  In Poster Session: Division D Section 1 Poster Session 2

Abstract

A Monte Carlo simulation study investigated full information maximum-likelihood estimation (FIML) performance in multilevel structural equation modeling (SEM) with missing data and different intra-class correlations (ICCs) coefficients. The study simulated the influence of missing data patterns and ICCs in multilevel SEM on five outcome measures (model rejection rates, parameter estimate bias, standard error bias, coverage, and power). Results indicated that FIML parameter estimates were robust for data missing on outcomes and/or higher-level predictor variables for data missing completely at random (MCAR) and data missing at random (MAR). While FIML estimation yielded relatively substantial decrease in parameter and standard error bias when data was not missing on higher-level variables, and in high rather than in low ICC conditions (.50 vs .20).

Author

• Chunling Niu, University of Incarnate Word