A Score-Based Method for Detecting Item Compromise and Preknowledge in CAT

• In Event: Computer Adaptive Testing Applications

Abstract

We develop a score-based method for detecting compromised items and examinees with preknowledge in computerized adaptive tests. A detailed simulation study reveals that the new method produces small false positive rates and large true positive rates across most conditions.

Authors

• Kylie Gorney, Michigan State University

• Chansoon Lee, American Board of Internal Medicine

• Jianshen Chen, COLLEGE BOARD