Bayesian Adaptive Lasso for Detecting Item-Trait Relationship and Differential Item Functioning in Multidimensional...

Area of Science:

• Psychometrics
• Statistical modeling
• Educational measurement

Background:

• Accurate identification of latent traits is essential in multidimensional tests.
• Differential item functioning (DIF) is critical for valid group comparisons but often studied separately from item-trait relationships.

Purpose of the Study:

• To develop a unified framework for detecting both item-trait relationships and DIF within multidimensional item response theory (MIRT) models.
• To integrate DIF effects into MIRT models as a variable selection problem for latent/observed variables and their interactions.

Main Methods:

• A Bayesian adaptive Lasso procedure was developed for simultaneous variable selection.
• This method allows for the concurrent estimation of item-trait relationships and DIF effects.

Main Results:

• Simulation studies demonstrated the method's effectiveness in parameter estimation.
• The procedure successfully recovered item-trait relationships and detected DIF effects.

Conclusions:

• The proposed unified framework and Bayesian adaptive Lasso procedure offer a robust approach for analyzing multidimensional item response data.
• This method enhances the accuracy of latent trait identification and ensures fairness across different groups by simultaneously addressing item-trait relationships and DIF.