Simplifying the Assessment of Measurement Invariance over Multiple Background Variables: Using Regularized Moderated...

Area of Science:

• Psychometrics
• Statistical Modeling
• Psychological Measurement

Background:

• Measurement invariance (MI) ensures assessment validity across groups.
• Traditional differential item functioning (DIF) methods struggle with complex, multi-variable analyses.
• Moderated Nonlinear Factor Analysis (MNLFA) offers a more comprehensive approach to MI/DIF.

Purpose of the Study:

• To develop a regularization approach for MNLFA to efficiently detect DIF.
• To address the limitations of conventional DIF detection methods in complex MNLFA models.
• To provide an automated and robust procedure for evaluating measurement invariance.

Main Methods:

• Proposed a regularization method for MNLFA estimation.
• Penalized the likelihood function for DIF parameters to promote sparsity.
• Evaluated the method through simulation studies and an empirical validation.

Main Results:

• The regularization approach effectively handles MNLFA estimation.
• The method successfully identifies differential item functioning across multiple variables.
• Demonstrated good performance in both simulation and real-world data.

Conclusions:

• Regularization offers a scalable and automated solution for DIF detection in MNLFA.
• This approach enhances the evaluation of measurement invariance across diverse populations.
• The proposed method improves the practical application of advanced psychometric techniques.