CAT and Item Misfit: Robustness, Consequences, and Practical Guidance

• In Event: AERA Poster Session Wednesday 11:45am
  In Poster Session: Poster Session I -- AERA Division D Section 1: Measurement, Psychometrics, and Assessment

Abstract

This study examines item misfit in computerized adaptive testing (CAT) through a PRISMA-guided systematic review and simulation. We identified inconsistent item fit evaluation practices, including reliance on indices unsuited for CAT. Simulation findings showed that CAT retained robust classification accuracy and stable theta estimates under misfit. An iterative purification method based on outfit and infit MSE values refined item pools but had limited impact on RMSE, suggesting CAT’s resilience when item banks are well-calibrated and population ability distribution is considered.

Authors

• Beyza Aksu-Dunya, Bartin University

• Stefanie A. Wind, University of Alabama