Effects of Practical Constraints on Item Selection Rules at the Early Stages of Computerized Adaptive Testing

摘要: The purpose of this study was to compare the effects four item selection rules—(1) Fisher information (F), (2) with a posterior distribution (FP), (3) Kullback-Leibler (KP), and (4) completely randomized (RN)—with respect precision trait estimation extent usage at early stages computerized adaptive testing. comparison rules carried out under three conditions: (1) using only function as criterion; both content balancing; function, balancing, exposure control. When test length less than 10 items, FP KP tended outperform F extreme levels in Condition 1. However, more realistic settings, it could not be concluded that outperformed F, especially when control imposed. greater nonrandom procedures performed similarly no matter what condition was, while had slightly higher usage.