A Comparison of Person-Fit Statistics in Computerized Adaptive Test Using Empirical Data From Two CAT Algorithms

• In Event: Roundtable Session 60
  In Roundtable Session: Computerized Adaptive Testing Issues

Abstract

One of the important issues considered in CAT is the aberrant response patterns caused by various reasons such as memorization (pre-knowledge of some items), random guessing, anxiety, fatigue, carelessness, etc. These aberrant response patterns usually cause the misfit between the data and the underline item response theory (IRT) model and results in the overestimation or underestimation of the ability. The purpose of this study is to compare CAT PFSes based on empirical data from a large scale CAT program with two different CAT algorithms. The data used in this study is the operational data collected from real examinees in a large-scale CAT program. This study would investigate the performance of the PFSes under various CAT algorithms and CAT settings.

Authors

• Chingwei D. Shin, Pearson

• Yuehmei Chien, Pearson