Evaluating Goodness-of-Fit Indices in Testing Measurement Invariance in Longitudinal Data

• In Event: Measurement Models

Abstract

The use of Goodness-of-fit indices (GFIs) has been investigated in testing Measurement Invariance (MI) between independent groups. Cut-off values (e.g., CFI=.01) associated with nominal Type I error were recommended. However, it is not known whether these cut-off values apply to longitudinal MI. This Monte Carlo study evaluates the use of some recommended GFIs in various MI configurations and data characteristics for longitudinal data. CFI, RMSEA, and normed Chi-square are robust to assorted lack of MI patterns and contamination levels whereas SRMR and model Chi-square are found to be sensitive to percent of items having non-invariant item parameters. Cut-off values are found for each GFI under examination in detecting longitudinal MI.

Authors

• Caiyan Zhang, The College Board

• Feifei Ye, University of Pittsburgh