DIF Analysis with Unknown Groups and Anchor Items.

Area of Science:

• Psychometrics
• Statistical modeling
• Educational measurement

Background:

• Ensuring fairness in surveys and tests is critical.
• Differential Item Functioning (DIF) analysis assesses item-level measurement invariance.
• Traditional DIF methods require known comparison groups and anchor items, which are often unavailable.

Purpose of the Study:

• To propose a general statistical framework for DIF analysis when both comparison groups and anchor items are unknown.
• To develop a method that simultaneously identifies latent subgroups and DIF items.
• To provide a computationally efficient algorithm for solving the proposed model.

Main Methods:

• A novel statistical framework modeling unknown groups via latent classes.
• Introduction of item-specific DIF parameters.
• An L1-regularized estimator to simultaneously identify latent classes and DIF items.
• A computationally efficient Expectation-Maximization (EM) algorithm for optimization.

Main Results:

• The proposed framework effectively handles DIF analysis without prior knowledge of groups or anchor items.
• Simulation studies demonstrate the method's performance.
• The approach was successfully applied to real-world educational test data.

Conclusions:

• The developed statistical framework offers a robust solution for DIF analysis in challenging scenarios.
• This method enhances the assessment of measurement invariance and fairness in educational and survey instruments.
• The findings contribute to advancing psychometric methods for detecting item bias.