Item-Focused Trees for the Detection of Differential Item Functioning in Partial Credit Models

Area of Science:

• Psychometrics
• Educational Measurement
• Statistics

Background:

• Differential item functioning (DIF) is crucial for fair assessments.
• Existing DIF detection methods primarily support binary response data.
• Methods for ordered response categories, like the partial credit model, are limited.

Purpose of the Study:

• To investigate and propose a novel method for detecting DIF in the partial credit model.
• To address the scarcity of DIF detection techniques for ordered response categories.
• To provide an accessible and interpretable approach for identifying DIF items.

Main Methods:

• Development of an item-focused tree methodology for DIF detection.
• Application of tree visualization to identify variables inducing DIF.
• Comparison with alternative DIF detection approaches.
• Use of simulations to evaluate the proposed method's performance.

Main Results:

• The proposed tree-based method effectively detects DIF in the partial credit model.
• Tree visualizations clearly illustrate the nature and source of DIF.
• The method offers an accessible alternative to existing techniques for ordered data.
• Simulations confirm the method's performance compared to alternatives.

Conclusions:

• The novel tree-based approach enhances DIF detection for the partial credit model.
• Visualizing DIF through trees improves understanding of item bias.
• This method offers a valuable tool for ensuring fairness in assessments with ordered response categories.