Comparing Model Fit Indexes Across Different Estimators in Structural Equation Modeling With Ordered Categorical Variables

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

Abstract

The application of RMSEA, CFI, and TLI highly relies on the conventional cutoff values developed under the normal-theory ML estimator in structural equation modeling. When data are ordered categorical, the ULS and DWLS estimators are recommended in practice. However, no clear guideline exists regarding the model fit indexes for these two estimators. This study compares RMSEA, CFI, and TLI based on ML with those based on ULS and DWLS. The purpose of this study is to answer: Given a population covariance matrix and a hypothesized model, if ML results in a specific RMSEA value (e.g., .06), what is the RMSEA value when ULS or DWLS is applied? CFI and TLI are investigated in a similar fashion.

Authors

• Yan Xia, Arizona State University - Tempe

• Yanyun Yang, Florida State University