欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
本文对CAT中能力估计的常用方法——最大似然估计法（MLE）进行改进,研究中结合EAP方法提出了改进的MLE法（R-MLE）。Monte Carlo模拟研究发现：不论是在定长CAT还是非定长CAT中,不论是在1PL模型下还是在2PL或3PL模型中,不论是在何种CAT题库结构下,R-MLE法较传统的MLE法具有更佳的估计精度及更有效的测验效率;R-MLE法不仅可以提高CAT的能力估计精度还可以进一步改善CAT测试的效率,具有一定的应用前景。
英文摘要：
In this paper,referring to the most popular ability estimation algorithm（ maximum likelihood estimation method, MLE）, some modification were done integrated into expected a posterior method（ EAP）, the new algorithm was called R - MLE method. The basic idea of this method was the following：once the score of the examinee was zero or full,his ability was estimated by EAP method;otherwise it was estimated by MLE method. Thus the adaptive choose of items was started from the second item in CAT,which was expected to be more effective and more adaptive than ever. The Monte Carlo simulation method was used here. The ABS index was used to test thepreeision of ability parameter estimate and the average use ration index of items was used to test the efficiency of testing. Two studies were employed here. The first one was designed to compare the precision of ability parameter estimation between R - MLE algorithm and MLE algorithm under the fixed and unfixed test length rule of CAT and under 1PLM ,2PLM and 3PLM. The second one was employed to compare the precision of ability parameter estimation between R - MLE algorithm and MLE algorithm under different structure of item bank, hut only the 2PLM being considered. The findings suggested ： Under any kind of CAT item bank, whether the fixed test length rule or unfixed test length rule was used, whether the one, two or three parameter logistic model were used, it was found that the estimation accuracy and efficiency of the R - MLE method was greater than that of the MLE method. It was also found that it would be more effective during CAT test when R - MLE method was used,which would be more applicable in practice.
同期刊论文项目
同项目期刊论文
期刊信息
位置：