Evaluating Fit Indexes for Multivariate t-Based Structural Equation Modeling

• In Event: 32.082 - Roundtable Session 11
  In Roundtable Session: 32.082-16 - Structural Equation Modeling With Difficult Data

Abstract

In conventional structural equation modeling (SEM), normal-theory maximum likelihood (ML-Normal) is not efficient and can be severely biased with the presence of even a tiny portion of outliers (or influential observations). On the other hand, the multivariate-t distributional family is a robust alternative to downweight the influence of outliers. Although multivariate-t maximum likelihood (ML-t) estimation has recently been made accessible in SEM software, to our knowledge its usage to handle outliers has not been shown in SEM literature. In this presentation we demonstrate the use of (ML-t) using the classic Holzinger and Swineford (1939) dataset. A simulation study is then conducted to examine the performance of fit-indexes and information criteria under ML-Normal and ML-t in the presence of outliers.

Authors

• Hok Chio (Mark) Lai, University of Cincinnati

• Jiaqi Zhang, SKT Education Group