Comparing Structural After Measurement Approaches for Structural Equation Models with Latent Interactions and Missing Data

• In Event: AERA Roundtable Session Sunday 11:45 am - Gold 4
  In Roundtable Session: Advanced SEM Approaches for Modeling Interaction, Heterogeneity, and Variability

Abstract

We compare various Structural After Measurement (SAM) estimation approaches for Structural Equation Models (SEMs) with latent interactions and missing data. Specifically, a SAM approach using a Croon based correction (SAM-Croon) has been shown to effectively accommodate latent interactions but missing data complicates the traditional factor score prediction on which it relies. In response, we extended SAM-Croon with alternative missing data handling techniques and used simulation studies to evaluate these methods for SEMs with single and multiple latent interactions and varying sample sizes and percent missingness. Results indicate that the optimal approach depends on model complexity, sample size, and percent missingness. This research provides crucial tools for educational researchers to analyze complex SEMs in the presence of missing data.

Authors

• Kyle T. Cox, University of North Carolina - Charlotte

• Benjamin Kelcey, University of Cincinnati

• Fangxing Bai, Montana State University