Rasch模型项目参数不变性的验证研究  

A Verification Study on Parameter Invariance of Rasch Model

作　　者：李付鹏 宋吉祥 杜海燕 Li Fupeng;Song Jixiang;Du Haiyan

机构地区：[1]安徽省教育招生考试院命题中心

出　　处：《教育测量与评价》2019年第1期18-24,39,共8页Educational Measurement and Evaluation

基　　金：2017年度国家教育考试科研规划课题(编号:GJK2017029)的研究成果之一

摘　　要：Rasch模型具有被试参数和项目参数相互独立的性质,即被试能力与项目难度无关。本研究以某年度大学入学考试数学学科的实测成绩数据为例,在随机抽样、不同性别抽样、不同水平群体抽样等条件下,对Rasch模型项目参数不变性进行了验证研究。研究表明:Rasch模型项目参数不变性验证的前提条件较为严格,需要排除诸多干扰因素的影响;Rasch模型项目参数不变性的验证存在一定的误差,无法实现与理论一致的"不变性";Rasch模型项目参数不变性没有统一的标准,需依据实际问题确定。The Rasch model has the property that the subject parameters and the item parameters are independent of each other,that is,the subject ability has nothing to do with the difficulty level of the project. Taking the measured data of mathematics of college entrance examination in a certain year as an example,this study conducted an empirical study on the itemparameters invariance of the Rasch model under the conditions of random sampling,different gender sampling,and sampling at different level groups. The research shows that the preconditions for Rasch parameter invariance verification are strict,which needs to exclude the influence of many interference factors;there are some certain errors in the verification of Rasch parameter invariance,and it is impossible to achieve the “invariance” consistent with the theory;there is no unified standard for Rasch parameter invariance,and it needs to be determined based on actual problems.

关 键 词：RASCH模型 参数不变性 DIF 抽样 

分 类 号：G40-051[文化科学—教育学原理]