Methodenzentrum

08.11.2022

New Publication: Bayesian penalty methods for evaluating measurement invariance in moderated nonlinear factor analysis. Brandt, H., Chen, S. M., & Bauer, D. J. (in press).

In this recent publication, we investigate how differential item functioning (DIF) can be investigated in large-scale settings with many covariates that may cause DIF, that is, responses to items do not only depend on the actual underlying construct of interest but also third variables. We show how standard methods such as using normal priors or small variance priors may fail to detect DIF, and that more sophisticated penalty priors such as spike-and-slab priors are necessary to identify the relevant parameters out of a huge parameter space.

 

Brandt, H., Chen, S. M., & Bauer, D. J. (in press). Bayesian penalty methods for evaluating measurement invariance in moderated nonlinear factor analysis. Psychological Methods.

Zurück