拟准检定法实施过程的有关问题  

Related questions on implementation of quasi-accurate detection of outliers

作　　者：王爱生[1] 欧吉坤[1] 赵长胜[2] 

机构地区：[1]中国科学院测量与地球物理研究所 [2]徐州师范大学国土信息与测绘工程系,江苏徐州221116

出　　处：《辽宁工程技术大学学报（自然科学版）》2006年第3期345-347,共3页Journal of Liaoning Technical University (Natural Science)

基　　金：中国科学院知识创新工程基金资助项目(KZCX2-106);江苏省教育厅的科研基金资助项目(03KJD420208)

摘　　要：为了使拟准检定法能够更加可靠、客观、实用,便于实现自动化,对拟准检定法进行深入的理论研究和数值试验的基础上,对观测值的排序指标、粗差的判别准则、判别标准中涉及的单位权中误差等有关实施细则做了适当的改进。当观测值较少时,初选之前观测值的排序指标宜采用单位权中误差进行;粗差的判别和复选之前的排序都应该考虑权重的影响;判别准则中的单位权中误差应该采用先验值。进而对改进后的拟准检定法的可靠性进行了大量的试验,试验结果表明该方法的成功率较高,可靠性较好。最后给出一个算例说明改进后的算法的检测过程。In order to make the QUAD （Quasi-Accurate Detection of Outliers）more reliable, objective, practicable and convenient to be automated, the authors modified observation order index, gross error determination standard, and unit weighted root mean square （rms） correlated to gross error determination standard. When the number of observation is small the order index for initial selection should be unit weighted rms; the weight should be take into account; the unit weighted rms correlated to gross error determination standard should adopt empirical value. Large number of reliability tests have been done andthe test result indicates that modified QUAD can detect outliers more successfully and credibly .At last a simulated example is presented, which demonstrates the process of monitoring outlier with this method.

关 键 词：拟准检定法 粗差 LEGE法 

分 类 号：TB9[一般工业技术—计量学]