CAT初始阶段被试能力估计方法改进探究

The Improvement of Estimators in the Initial Ability Estimation of CAT

在CAT初始阶段被试作答的信息较少,对被试能力进行合理和稳健的估计比较困难。本文设计CAT测试过程的代表性测验样例,使用不同能力估计方法比较CAT初始阶段的被试能力估计情况。在单参数模型下,在题量较少时,EAPE-N(0,0.5)方法的能力估计的合理性和稳健性要相对优于MLE、EAPE-N(0,2)、EAPE-N(0,1)方法。在CAT测试过程中可以先使用先验分布方差较小的EAPE方法进行估计,随着测试题量增加,逐步增大EAPE方法的先验分布方差,并可以和a分层选题策略结合起来使用。

From initial ability estimation to interim ability estimation in the process of a computerized adaptive test,there are a few responses to the items. The authors design a classic sample test which represents a part of CAT in the initial ability estimation. Under one-parameter logistic model,the ability estimation of the method of EAPE-N（0,0.5） is more accurate and robust than the method of MLE,EAPE-N（0,1）,EAPE-N（0,2）. The authors propose new methods that in the process of ininitial ability estimation and interim ability estimation：firstly we can use the method of EAPE with smaller prior variance,for example,EAPE-N（0,0.5）. As the examinees make responses to more items,we can use the method of EAPE with bigger prior variance,for example,EAPE-N（0,1.5）,or EAPE-N（0,2）. And a-Stratified multistage can be applied in the initial estimation while the method of the adjusted EAPE is uosed.