Two-Stage Uniform Adaptive Testing to Balance Measurement Accuracy and Item Exposure
Two-Stage Uniform Adaptive Testing to Balance Measurement Accuracy and Item Exposure
复制标题
两阶段统一自适应测试以平衡测量精度和项目暴露
DOI：
10.1007/978-3-031-11644-5_59
复制
发表时间：
2022
期刊：
影响因子：
--
通讯作者：
Yoshimitsu Miyazawa
中科院分区：
文献类型：
--
作者：
M. Ueno;Yoshimitsu Miyazawa
计算机化自适应测试（CAT）提出了一个权衡问题，涉及增加测量精度与减少项目暴露在一个项目池。为了解决这个困难，我们提出了两阶段的统一自适应测试。在第一阶段中，所提出的方法分区一个项目池成许多统一的项目组使用一个国家的最先进的统一测试组装技术的基础上的随机可编程最大团问题。然后从一个统一的项目组中选择最优项目。在第二阶段中，当被试能力估计的标准误差变得小于一定值时，它切换到从整个项目库中选择并呈现最佳项目。数值实验结果表明了该方法的有效性。
Computerized adaptive testing (CAT) presents a tradeoff problem involving increasing measurement accuracy vs. decreasing item exposure in an item pool. To address this difficulty, we propose two-stage uniform adaptive testing. In the first stage, the proposed method partitions an item pool into numerous uniform item groups using a state-of-the-art uniform test assembly technique based on the Random Integer Programming Maximum Clique Problem. Then the method selects the optimum item from a uniform item group. In the second stage, when the standard error of an examinee’s ability estimate becomes less than a certain value, it switches to selecting and to presenting an optimum item from the whole item pool. Results of numerical experiments underscore the effectiveness of the proposed method.