The Sensitivity of Bayesian Fit Indices to Structural Misspecification in Structural Equation Modeling Using Bayesian Estimation

• In Event: Data Analysis and Statistics

Abstract

Global fit indices, such as RMSEA, CFI, and TLI were not sensitive to structural misspecification in structural equation modeling (SEM). However, it remains unknown about the sensitivity of their Bayesian variants to structural misspecification. This study examined the performance of PPP, DIC, BRMSEA, BCFI, BTLI with varying sample sizes, model size, the severity of structural misspecification, and choice of priors. The results indicated that BRMSEA, BCFI, and BTLI were not sensitive to structural misspecification in Bayesian SEM, similar to their frequentist counterparts. However, the sensitivity of PPP and DIC depended on the design factors. The magnitude of factor loadings had an positive impact on the sensitivity of PPP and DIC.

Authors

• Chunhua Cao, University of Alabama
  Presenting Author

• Benjamin Lugu, University of Alabama
  Presenting Author

• Jujia Li, University of Alabama
  Presenting Author

• Kaiwen Man, University of Alabama
  Non-Presenting Author