基于Bayes估计和Rasch模型拟合的考试缺失数据分析

发布时间：2018-02-28 12:04

本文关键词： Bayesian估计 Rasch模型缺失数据　出处：《贵州师范大学》2017年硕士论文　论文类型：学位论文

【摘要】：研究以贵州省黔南州某高中2000名高三学生英语测试的9道单项选择题结果为原始数据,运用Matlab和Winsteps软件对数据进行随机缺失Bayes估计处理和Rasch模型拟合。得出结论:第一,本研究使用Gibbs抽样对英语测试数据中的缺失数据进行处理比用一般均值替代法处理的缺失值对照组更接近原始数据真实的参数估计结果,证明了缺失数据合理处理的有效性和必要性。第二,本研究通过处理不可忽略的缺失数据能更真实反应英语测试试题的特性,证明了合理处理缺失数据的合理性与科学性。第三,本研究所选取的英语测试单项题目各个统计指标均符合心理测量学标准,题目设置相对合理,确实测量的是被试的英语水平这一特质。
[Abstract]:Based on the results of 9 multiple choice questions of 2 000 senior middle school students in a high school in Qiannan Prefecture of Guizhou Province, the study uses Matlab and Winsteps software to process the randomly missing Bayes estimation and Rasch model fitting. The conclusions are as follows: first, This study uses Gibbs sampling to process the missing data in English test data, which is closer to the real parameter estimation result of the original data than the control group treated by the general mean substitution method. It proves the validity and necessity of the reasonable processing of missing data. Secondly, this study can more truly reflect the characteristics of English test questions by processing the missing data which cannot be ignored. It proves that it is reasonable and scientific to deal with the missing data reasonably. Thirdly, each statistical index selected in this study accords with the standard of psychometrics, and the problem setting is relatively reasonable. It is true that the English proficiency of the subjects is measured.
【学位授予单位】：贵州师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：B844.2

【参考文献】