3V-JMLE法:计算机自适应测试中一种基于三值矩阵的联合参数估计方法*

摘要

In this paper, the parameter estimation method commonly used in computer adaptive testing, the joint maximum likelihood estimation (JMLE) method, was improved. Meanwhile, a joint parameter estimation method based on three-valued matrix, abbreviated as 3V-JMLE method, was proposed. Based on this, the paper took the tested answer information in the answer response database as the sample, and used the JMLE and 3V-JMLE methods to conduct the parameter estimation, respectively. The comparison results showed that JMLE method and 3V-JMLE method had same level of parameter estimation accuracy and calculation efficiency under the ideal answer matrix. However, under the non-ideal answer matrix, the JMLE method had certain limitations in the parameter estimation, while the 3V-JMLE method had high parameter estimation accuracy and greatly improved the calculation efficiency. The proposal of 3V-JMLE method had important guiding significance for the actual estimation parameter process of the joint parameter estimation method.%文章对计算机自适应测试中常用的参数估计法——联合极大似然估计(Joint Maximum Likelihood Estimation,JMLE)法进行了改进,提出了一种基于三值矩阵的联合参数估计方法——3V-JMLE 法.基于此,文章以作答反应数据库中被试作答信息为样本,分别采用JMLE法、3V-JMLE法进行参数估计,其对比结果表明:在理想作答矩阵下,3V-JMLE法和JMLE法具有同等的参数估计精度和计算效率;在非理想作答矩阵下,采用JMLE法进行参数估计存在一定的局限性,而采用3V-JMLE法具有很高的参数估计精度并大大提高了计算效率.3V-JMLE法的提出,对于联合参数估计方法的实际估计参数过程有重要指导意义.