Bayesian Analysis of Longitudinal Approximate Measurement Invariance in Autoregressive Cross-Lagged Models

• In Event: 18.047 - Statistical Research in the Contexts of Propensity Scores, Longitudinal Models, and Bayesian Analyses

Abstract

To accurately infer longitudinal relationships among latent factors, longitudinal measurement invariance (MI) needs to be assumed. Alternative to placing equality constraints on parameters, approximate MI can be assessed by specifying informative priors on parameter differences between time points. This study examined the impact of longitudinal approximate MI in autoregressive cross-lagged models using Bayesian structural equation modeling (BSEM). Design factors included sample sizes, factor structures, conditions of non-invariance, and magnitudes of structural coefficients. Correctly specified models were analyzed using BSEM. Model fit and structural parameters were evaluated. Preliminary results show that with mixed patterns of non-invariance, even though the model fitted well, structural parameter estimates may be biased especially with small sample sizes and relatively large prior variances.

Authors

• Xinya Liang, University of Arkansas

• Yanyun Yang, Florida State University

• Jiajing Huang, Florida State University