计算机自适应测验中Rasch模型稳健性的模拟研究  

A Simulation Study of Rasch Model's Robustness in Computerized Adaptive Test

作　　者：邓远平[1] 蔡艳[2] 罗照盛[1] 

机构地区：[1]江西师范大学教育学院 [2]江西师范大学数信学院

出　　处：《考试研究》2006年第3期87-94,共8页Examinations Research

摘　　要：本研究采用模拟数据的方法,在计算机自适应测验(Computer Adaptive Test,简称CAT)中分别采用Rasch及Birnbaum两种模型估计能力,通过比较两者的误差均方根(Root Mean Square Error,简称RMSE)、平均差异(Average Deviation,简称AD)及能力相关,对Rasch模型在CAT中的稳健性进行了研究。结果发现Rasch模型在区分度不等的条件下仍然能较准确地估计被试的能力水平,具有很强的稳健性。Computerized adaptive test (CAT) is based on Item response theory, select item according to some rule of Item-selecting and examinees responses to items to estimate the value of ability accurately The paper usesthe method of simulating data, estimate the value of ability with Rasch model and Birnbaum model respectively in the CAT, study the Rasch model s robustness through comparing Root Mean Square Error(RMSE) and Average Deviation (AD) and correlations of ability between Rasch model and Birnbaum model, the outcome finds the Rasch model in CAT is robust even if Rasch Model' s assumption (equal item discrimination parameters) can t be satisfied.

关 键 词：CAT RASCH模型 Bimbaum模型 稳健性 

分 类 号：TP306.2[自动化与计算机技术—计算机系统结构]