剔除离群值的学生化残差新方法  

The Novel Method for Outliers' Rejection of the Studentized Residual Error

作　　者：王中宇[1] 张海滨[2] 刘智敏[3] 

机构地区：[1]北京航空航天大学仪器科学与光电工程学院,北京100083 [2]空军装备设计研究院情报所,北京100085 [3]中国计量科学研究院,北京100013

出　　处：《仪器仪表学报》2006年第6期624-628,637,共6页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(50375011);北京市自然科学基金(3043019)资助项目

摘　　要：在计量测试领域中,离群值就是含粗差即粗大误差或过失误差的测量值。首先研究了学生化残差及其绝对值的基本概念、性质及其分布函数,提出了剔除离群值的学生化残差新方法,该方法不需要区分离群值在上侧或下侧的情况,因此所得到的判断条件更加方便。采用蒙特卡洛模拟法求出了测量次数小于50时,各分布函数、均值及其标准差的具体数值。最后得到了剔除离群值的学生化残差绝对值法的临界值表,并通过测量实例验证了理论分析的正确性,与现有的判别方法进行比较,结果表明该方法可以迅速、可靠地剔除测量数据列的离群值。The outliers are the measured values including gross errors or fault errors in the metrology field. The general concepts of the studentized residual error and its absolute value, qualities and distribution functions are introduced. The absolute value method for rejecting the outliers of studentized residual error is also put forward. It is unnecessary to distinguish the upside or downside of the outliers and therefore the judgement condition is more accurate. The detailed value of each distribution function, mean and standard deviation are calculated by Monte Carlo simulation when the measuring number is within 50. Finally the critical value table is obtained for rejecting the outliers of the studentized residual error. The results demonstrated the effectiveness and consistency of the theoretical assessment by the measurement experiments. This method can be used to reject the outliers in the measurement sequence in a simple, rapid and reliable way by comparison with the current methods.

关 键 词：离群值 剔除 学生化残差 蒙特卡洛模拟 

分 类 号：TB9[一般工业技术—计量学]