A structural after measurement approach to bifactor predictive models

The bifactor model has regained popularity due to its conceptual appeal. However, the bifactor predictive model, which extends a bifactor model to a criterion variable, often encounters empirical under-identification due to approximate linear dependency. This limitation leads to various statistical issues (e.g., non-convergence, estimation bias, inaccurate standard errors), hindering the use of the bifactor model for predictive purposes. To address this limitation, we introduced the recently developed Structural After Measurement (SAM; Rosseel & Loh, 2022) approach to the bifactor predictive model and examined its robustness with a series of Monte Carlo simulations. Our simulation results indicated that the SAM approach effectively enhances the statistical performance of bifactor predictive models compared to the SEM approach in terms of model convergence, stability of point estimates, accuracy of standard error estimates, coverage rates, and Type I error rates, at the cost of slight bias. Our empirical illustration also supported the simulation findings, further illustrating the effectiveness of the SAM approach.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Psychology

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2024

Author dc:creator

• Choi, Jinsoo

Contributors dc:contributor

• Zhang, Bo

• Structural After Measurement
• Bifactor Model
• Augmentation

Statement dc:rights

• Copyright 2024 Jinsoo Choi

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/124626

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F124626