基于Rasch模型的参数估计方法比较研究  

The Comparison between the Method of MLE, MLE/EM and BMES under the Rasch Model

作　　者：王佶旻[1] 李潇[2] WANG Jimin LI Xiao(Beijing Language and Culture University, Beijing 100083, China Beijing Normal University, Beijing 100875, China)

机构地区：[1]北京语言大学,北京100083 [2]北京师范大学,北京100875

出　　处：《中国考试》2017年第9期11-21,共11页journal of China Examinations

基　　金：北京市社科规划项目"首都留学生跨文化适应研究"(项目号:13WYB014);北京语言大学院级项目(项目号:17YJ050011)的资助

摘　　要：本研究的目的是基于Rasch模型,比较联合极大似然估计法、边际极大似然估计法和EM算法、边际贝叶斯估计法参数估计结果的准确程度。实验数据为2 185名被试在HSK试卷170道试题中的作答矩阵,考虑到初值和收敛精度对参数估计结果的影响,将三种参数估计方法按照初值设置和收敛精度不同分别进行参数估计,然后通过计算项目参数估计标准误判断参数估计方法的准确度。The objective of this article is to assess the accuracy of the Joint Maximum Likelihood Estimation（JMLE）, the Marginal Maximum Likelihood Estimation/EM algorithm（MMLE/EM）, Marginalized Bayesian Parameter Estimation（BMEs）based on the single parameter logistic model. Experimental subject is the answer matrix of 2 185 examinees, who were tested in the HSK examination（including 170 questions）on Dec 9th, 2005.We assessed the accuracy of parameter estimation by comparing the standard errors from each estimation method.This study also conducted the parameter estimation under different initial values and convergence precisions for each approach, taking into account the effects of initial value and convergence precision setting for parameter estimation results.

关 键 词：项目反应理论 参数估计 联合极大似然估计法 边际极大似然估计法和EM算法 边际贝叶斯估计法 

分 类 号：G405[文化科学—教育学原理]