A Robust Method for Detecting Item Misfit in Large-Scale Assessments.

Area of Science:

• Psychometrics
• Statistical modeling
• Educational measurement

Background:

• Accurate scale construction requires identifying item misfit and Differential Item Functioning (DIF).
• Existing methods often assume perfect model-data fit, which can be unrealistic.
• Classical test theory and item response theory rely on explicit assumptions about item function.

Purpose of the Study:

• To develop a robust approach for detecting DIF that does not require perfect model-data fit.
• To provide a more reliable method for assessing item fit in scale construction.

Main Methods:

• Utilized Tukey's concept of contaminated distributions.
• Employed robust outlier detection techniques.
• Flagged items with inadequate model-data fit.

Main Results:

• Successfully identified items with inadequate model-data fit.
• Demonstrated a robust approach to DIF detection.
• Provided a method less reliant on idealized statistical assumptions.

Conclusions:

• The proposed robust method enhances the accuracy of DIF detection.
• This approach offers a more practical solution for scale construction and measurement.
• It addresses limitations of traditional methods by not assuming perfect model fit.