A Simulation Study Comparison of Estimating Structural Relationships With Small N in SEM

• In Event: Methodological Innovations in Structural Equation Modeling

Abstract

Traditional SEM might have convergence and accuracy issues to estimates the parameters in a model with small sample sizes with a relatively large number of items. There have been several different approaches for addressing this concern, for example, Bayesian SEM, item parceling, structural-after-measurement (SAM). This study focuses on comparing the performance of the above mentioned approached in estimating the structural relationship between latent factors. The preliminary results of using item parceling shows that the estimates of structural paths are biased, especially when standardized factor loadings are low. The complete results can shed light on the choice of approaches to addressing small sample size in SEM.

Authors

• Chunhua Cao, University of Alabama

• Christopher Adah Ocheni, University of Alabama