基于半参数平差模型的粗差定位与定值研究

发布时间：2018-01-27 13:48

本文关键词： 半参数模型估计值观测数据混合Cook距离粗差　出处：《武汉大学学报(信息科学版)》2016年11期 　论文类型：期刊论文

【摘要】：对于包含系统误差和粗差的观测数据,本文将混合Cook距离引入到半参数模型中,实现了粗差的定值定位。首先,构造补偿最小二乘函数,利用泰勒展式,根据均值漂移模型与删失模型的等价性,导出了观测数据中剔除第i个观测数据前后参数分量和非参数分量相应估计值之间的关系式,为粗差的定位分析奠定了基础。其次,将混合Cook距离作为诊断统计量,进行粗差定位分析,得到了参数分量和非参数分量的诊断统计量的简洁计算公式,为了提高粗差定位的准确性,给出了混合Cook距离参数Q和C的一些常用形式,通过合理选择相应参数,计算参数分量和非参数分量的距离函数,实现粗差的定位,并将系统误差和粗差从观测数据中区别开来。最后通过模拟算例和实测数据验证了本文方法的正确性。
[Abstract]:For the observation data including system error and gross error, the mixed Cook distance is introduced into the semi-parametric model to locate the gross error. Firstly, the compensated least square function is constructed and the Taylor expansion is used. According to the equivalence between the mean drift model and the censored model, the relationship between the corresponding estimations of the parameter components and the non-parametric components before and after the removal of the first observation data from the observed data is derived. It lays a foundation for the positioning analysis of gross error. Secondly, the mixed Cook distance is used as diagnostic statistic, and the simple calculation formula of diagnostic statistics of parameter component and non-parametric component is obtained. In order to improve the accuracy of gross error location, some common forms of mixed Cook distance parameters Q and C are given, and the distance functions of parameter components and nonparametric components are calculated by reasonably selecting the corresponding parameters. The gross error location is realized and the system error and gross error are distinguished from the observed data. Finally, the correctness of this method is verified by the simulation example and the measured data.
【作者单位】：中国地质大学信息工程学院;国家知识产权局专利局专利审查协作四川中心;
【基金】：国家自然科学基金(41374017,11471105,40974002)~~
【分类号】：P207
【正文快照】： 项目资助:国家自然科学基金(41374017,11471105,40974002)。随着观测手段的不断进步,观测数据中出现粗差和系统误差的概率增大,使得参数估计的精度受到严重的影响。在观测数据包含系统误差时,半参数模型能较好地将系统误差分离出来,得到较高精度的参数估计结果[1-5],当观测数

【相似文献】