两种新的计算机化自适应测验在线标定方法

发布时间：2018-02-03 10:13

本文关键词： 全功能极大似然估计计算机化自适应测验项目反应理论在线标定题库建设　出处：《心理学报》2016年09期 　论文类型：期刊论文

【摘要】：在线标定技术由于具有诸多优点而被广泛应用于计算机化自适应测验(CAT)的新题标定。Method A是想法最直接、算法最简单的CAT在线标定方法,但它具有明显的理论缺陷——在标定过程中将能力估计值视为能力真值。将全功能极大似然估计方法(FFMLE)与"利用充分性结果"估计方法(ECSE)的误差校正思路融入Method A(新方法分别记为FFMLE-Method A和ECSE-Method A),从理论上对能力估计误差进行校正,进而克服Method A的标定缺陷。模拟研究的结果表明:(1)在大多数实验条件下,两种新方法较Method A总体上可以改进标定精度,且在测验长度为10的短测验上的改进幅度最大;(2)当CAT测验长度较短或中等(10或20题)时,两种新方法的表现与性能最优的MEM已非常接近。当测验长度较长(30题)时,ECSE-Method A的总体表现最好、优于MEM;(3)样本量越大,各种方法的标定精度越高。
[Abstract]:Because of its many advantages, online calibration technology has been widely used in the computerized adaptive test (CAT) of the new problem calibration. Method A is the most direct idea. The simplest on-line calibration method for CAT is proposed. However, it has obvious theoretical defects: in the process of calibration, the capability estimate is regarded as the true value of capability. The full function maximum likelihood estimation method (FFMLEE) and the "use of adequacy result" estimation method (. The method of error correction is integrated into Method A (the new method is called FFMLE-Method A and ECSE-Method A respectively). The error of capability estimation is corrected theoretically to overcome the calibration defect of Method A. the simulation results show that the error of capacity estimation is in most experimental conditions. Compared with Method A, the two new methods can improve the calibration accuracy, and have the largest improvement on the short test with the test length of 10. (2) when the length of CAT test is short or medium (10 or 20), the performance of the two new methods is very close to that of the MEM with the best performance, and when the length of the test is longer than 30 questions). The overall performance of ECSE-Method A is the best, which is better than that of ECSE-Method; The larger the sample size, the higher the calibration accuracy of various methods.
【作者单位】：北京师范大学中国基础教育质量监测协同创新中心;
【基金】：国家自然科学基金青年基金项目(31300862) 高等学校博士学科点专项科研基金项目新教师类(20130003120002) 东北师范大学应用统计教育部重点实验室开放课题(KLAS 130028614)资助
【分类号】：B841
【正文快照】： 1引言量新的技能类型(如知识状态);(4)与多级项目反应理论(Polytomous Item Response Theory,PIRT)结合在传统纸笔测验(Paper-and-Pencil,PP)中,可以提供基于表现的题目类型(如开放题);(5)与多所有被试不论能力高低都作答相同的一批题目,所维IRT(Multidimensional IRT,MIRT)

【相似文献】