Croon’s Bias-corrected Estimation for Multilevel Structural Equation Models with Latent Interactions

Abstract

Consideration of moderation effects is important in a range of substantive applications that include nested data and latent variables. A critical limitation in multilevel structural equation models with computationally demanding latent interactions is the poor performance (e.g., convergence failure and biased estimates) of common estimators such as maximum likelihood. To address this issue, we developed extensions to Croon’s bias-corrected estimation for multilevel structural equation models with latent interactions. After detailing these new corrections, we conducted several simulation studies to probe the accuracy and efficiency of the estimator. It produced accurate coefficient estimates with a variety of latent interactions while avoiding major convergence failure issues even when sample sizes were limited. The availability of Croon’s estimation for two-level structural equation models with latent interactions provides researchers a viable alternative estimator even when sample sizes are limited.