基于项目反应理论的影响计算机化自适应测试a分层法的因素研究  

STUDY ON FACTORS IMPACTING a-STRATIFIED METHOD OF CAT BASED ON ITEM RESPONSE THEORY

作　　者：姚敏华[1] 王东[1] 

机构地区：[1]上海交通大学软件学院,上海200240

出　　处：《计算机应用与软件》2011年第10期149-154,173,共7页Computer Applications and Software

摘　　要：计算机化自适应测验CAT(Computerized Adaptive Testing)中的选题策略,一直是国内外相关学者关注的问题。选题策略是计算机自适应测验研究的一项重要内容,它的好坏直接关系到考试的信度、效度及考试的安全性。采用计算机模拟程序对影响a分层选题策略的各因素进行研究。得出了包括测试的终止条件应设为已测项目的信息总量或题目总数达到事先预定标准,层次划分数应为15层,各层信息量所占比例应逐渐上升,项目难度参数应服从均匀分布等各因素的最佳状态,实验证明,优化了a分层选题策略,能显著提高计算机自适应测验的准确性和稳定性。Selection strategy in computerised adaptive testing（CAT） has always been the concerned topic of domestic and foreign scholars.Selection strategy is an important part of the study of CAT.Its performance directly affects the reliability,validity and security of the test.In the paper the factors impacting the selection strategy of a-stratified selection strategy are studied using computer simulation programs.The best states of various factors derived are： the terminal condition of the test shall be set to that the gross information of the tested projects or the total number of the subjects meet the predetermined criteria;the partition number of the layer shall be fifteen;the proportion of the amount of information to each layer shall be escalated;and the difficulty degree of the project parameters shall be subjected to the uniform distribution.Experiments show that when the a-stratified selection strategy is optimised,the accuracy and stability of CAT can be significantly improved.

关 键 词：项目反应理论(IRT) 计算机化自适应测验 a分层选题策略 

分 类 号：TP302.2[自动化与计算机技术—计算机系统结构]