基于处理效应模型的评估方法改进及应用研究  

The Improvement and Application of Evaluation Method Based on Treatment Effect Model

作　　者：张建军[1] 乔松珊 吕海燕[1] ZHANG Jian-jun;QIAO Song-shan;LV hai-yan(Collage of Information and Management Science of Henan Agricultural University,Zhengzhou 450002,China;Department of public education of Zhengzhou University of Economics and Business,Zhengzhou 451191,China)

机构地区：[1]河南农业大学信息与管理科学学院,河南郑州450002 [2]郑州经贸学院公共教学部,河南郑州451191

出　　处：《数学的实践与认识》2021年第13期95-102,共8页Mathematics in Practice and Theory

基　　金：河南省科技攻关项目(212102311005);国家自然科学基金资助项目(31701309)。

摘　　要：处理效应模型广泛应用于教育、医学和公共政策研究中,模型估计结果依赖于处理对象的选择,特别是当样本量较小时,估计结论缺乏稳健性.在这种情形下,排序集抽样是一种经济有效的方法.基于排序集样本研究了改进的平均处理效应估计模型,对改进后估计量的无偏性和渐进性进行了证明,建立了估计量精度与回归系数估计精度之间的关系.结果表明,排序集抽样可以有效较好提高了平均处理效应估计的准确性.最后借助随机模拟和算例验证结论可靠性.The treatment effect model is widely used in education,medicine and public policy research,The estimation of the model depends on the selection of the treatment object.Especially small sample conditions,the estimation results are not robust.Under such circumstances,The ranked set samples is economical and effective method.The paper study the improvement of the treatment effect estimation based on the ranked set samples.The unbiased and asymptotic properties of the adjusted estimator are discussed.This analysis enables us to highlight the connection between the precision of the treatment effect estimate and the predictive accuracy of the regression coefficient.The results show that the ranked set samples can improve the accuracy of estimation.Finally The results are supported by stochastic simulation and practical example.

关 键 词：平均处理效应 排序集抽样 简单差分 回归 估计 

分 类 号：O212.1[理学—概率论与数理统计]