CAT选题策略研究

New Exploration of CAT Selection Strategy

传统CAT选题策略是将被试者的估计值作为真实值去选题目,使用的都是间接指标,旨在寻找最有可能的解。其缺点是计算方法复杂,误差最大值不易控制。针对传统CAT选题策略的特殊性:不离散,被估能力值不能直接求解,稳定性差,提出一种将能力值线性离散化,并使用"最大期望判准率"的方法来计算每次被试者的估计能力,旨在降低计算的难度,提高测验的精度及其稳定性。实验结果表明,该方法具有良好的性能。

Traditional CAT selection strategy is the estimate of the subjects as a real value to choose the best subject, using indirect index, looking for the most likely solution as purpose. But these methods have a common defect including complex calculation, maximum error controlled unwell. Based on the particularity of traditional CAT selection strategy such as no discrete, estimated value indirectly solving, poor stability, we put forward estimated value linear discretization, and use the ＂maximum expected rate of criteria ＂ to calculate subjects estimate value each time. It can increase test precision, reduce the difficulty of calculation and improve the stability of the test. The experimental results show that this method shows good performance.