A Comparison of Item Selection Rules at the Early Stages of Computerized Adaptive Testing

Abstract

The effects of five item selection rules—Fisher information (FI), Fisher interval information (FII), Fisher information with a posterior distribution (FIP), Kullback-Leibler information (KL), and Kullback-Leibler information with a posterior distribution (KLP)—were compared with respect to the efficiency and precision of trait (Θ) estimation at the early stages of computerized adaptive testing (CAT). FII, FIP, KL, and KLP performed marginally better than FI at the early stages of CAT for Θ= -3 and -2. For tests longer than 10 items,there appeared to be no precision advantage for any of the selection rules.