ISSN 0439-755X
CN 11-1911/B

中国科学院心理研究所

• 论文 •

### 二参数逻辑斯蒂模型项目参数的估计精度

1. (1西南大学数学与统计学院, 重庆 400715) (2重庆市教育考试院, 重庆 401147)
• 收稿日期:2012-12-07 出版日期:2013-10-25 发布日期:2013-10-25
• 通讯作者: 杜文久
• 基金资助:

重庆市教育科学规划项目(2011KS035)资助。

### The Item Parameters’ Estimation Accuracy of Two-Parameter Logistic Model

DU Wenjiu;ZHOU Juan;LI Hongbo

1. (1 School of Mathematics and Statistics, Southwest University, Chongqing 400715, China) (2 Chongqing Educational Examination Authority, Chongqing 401147, China)
• Received:2012-12-07 Online:2013-10-25 Published:2013-10-25
• Contact: DU Wenjiu

Abstract: The item parameters’ estimation accuracy is very important to test development, especially to the construction of item bank. Nowadays, in domestic and foreign, the study of item parameters’ estimation accuracy is mostly based on that the true values of the item parameters are known, under this circumstances, applying kinds of estimation methods to estimate new estimators, then evaluating methods’ effectiveness by comparison with the true values about the root-mean-square-error (RMSE) and Bias. But this method can’t present the change rules of estimation error among the different parameters’ true values. In order to solve this problem, this article attempts to explore the item parameters’ estimation accuracy issue on the point of item parameters estimation information function. Based on the two-parameter Logistic model, given the definition of the item parameters’ estimation information function firstly, then generated examinee sample to explore the factors which affected item parameters’ estimation accuracy by computer simulation randomly. The experiments were designed (2 × 3 × 2) cases. The study is conducted and revealed: (1) The accuracy of the estimators enhance as the quantity of the sample increases. (2) The examinee’s ability distribution has more influences on difficulty parameter estimation accuracy, while less influences on discrimination parameter. (3) The estimation accuracy of item discrimination and item difficulty parameter are both influenced by parameter a and parameter b’s true value.