小样本情况下试题Logistic IRT参数估计研究综述  

Researches on IRT Item Parameter Estimation under Small Sample Sizes:A Literature Review

作　　者：汪存友[1] 

机构地区：[1]山西师范大学教育技术与传媒学院

出　　处：《考试研究》2014年第6期48-60,共13页Examinations Research

基　　金：全国教育科学"十二五"规划教育部考试专项课题"人工神经网络在小样本试题IRT参数估计中的应用研究(GFA111008)"

摘　　要：回顾国内外有关小样本情况下估计试题的Logistic IRT参数的研究,可以总结出六种参数估计方法,分别是:修改IRT模型法、提供先验信息法、人工神经网络法、非参数估计法、经典测验理论标准化法以及使用数据增强技术。后续研究应加强对已有参数估计方法的改进,使用包括标准误在内的多种误差指标,在250人以内的样本水平上,采用模拟数据与真实数据相结合的模拟实验法开展更加严谨的模拟研究。Researches on estimating item parameters in Logistic IRT model under small sample sizes were reviewed in this paper, and six methods were found out, including modified IRT model approach, prior information approach, artificial neural network approach, nonparametric IRT estimation approach, standardized statistics of Classical Test Theory approach, and data augmented technique approach. Several aspects of the above approaches were analyzed and discussed, such as the principles, methods, conclusions, error indications and sample sizes. Future researches should pay more attention to the improvement of those item parameter estimation approaches, by applying simulation experimental method combining both simulated and realistic data, using more error indications, including standard error of estimations, when examinee sample size under 250.

关 键 词：小样本 项目反应理论 试题参数估计 

分 类 号：G427.74[文化科学—课程与教学论]