Evaluating CAT-adjusted Approaches for Suspected Item Parameter Drift Detection

Kevin Cappaert, Wen Yao, Yu-Feng ChangMeasurement 2018 · DOI

Abstract

Events such as curriculum changes or practice effects can lead to item parameter drift (IPD) in computer adaptive testing (CAT). The current investigation introduced a point- and weight-adjusted D2 method for IPD detection for use in a CAT environment when items are suspected of drifting across test administrations. Type I error and power rates of the proposed method were compared to a more traditional, non-adjusted D2 method and two recently suggested methods for use in a CAT environment: pseudo-count robust z and pseudo-count D2 methods. Though all CAT-adjusted IPD detection methods compared resulted in high power, the pseudo-count D2 method was found to have the highest power rates, with the proposed D2 method a close second. All four methods were found to have acceptable type I error rates.
BibTeX
@inproceedings{cappaert2018evaluating,
  author = {Kevin Cappaert and Wen Yao and Yu-Feng Chang},
  title = {{Evaluating CAT-adjusted Approaches for Suspected Item Parameter Drift Detection}},
  booktitle = {Measurement},
  year = {2018},
  doi = {10.1080/15366367.2018.1511199},
}

Loading related content...

Figure