由于一个观测值的粗差将影响多个观测值的残差估值，对双因子抗差估计的迭代计算方法作如下改进：当最大标准化残差的检验结果为粗差或者明显是可疑粗差时，只对粗差观测值构建双因子等价权，其他观测值的权不变。除上述两种情况外，可同时对所有观测值构建双因子等价权进行抗差估计，并基于改进的双因子等价权构建方法简化计算步骤。利用GPS控制网的观测数据进行模拟计算分析表明，该改进算法在含有明显粗差时可提高参数估值的精度；在粗差不明显或个数较少时，可在保持参数估计精度的基础上提高计算效率。"/>
Since the mass of one observation will affect the residual of multiple observations, this paper improves the iterative method of bifactor robust estimation as follows: if the maximal standardized residual is tested as mass or obvious suspicious mass, the bifactor equivalent weight is only constructed for the observation with mass, while the weight of other observations remains unchanged. In addition to the two cases above, the bifactor equivalent weight can be constructed for all observations at the same time. The calculation steps can also be simplified based on the improved equivalent weight construction method. Simulation calculation and analysis are carried out using the observation data of a GPS network. The results demonstrate that the method can improve calculation efficiency when containing a small amount of masses, and improve the accuracy of parameter estimation when containing more masses."/>
相关观测值双因子抗差估计的改进算法
Abstract：Since the mass of one observation will affect the residual of multiple observations, this paper improves the iterative method of bifactor robust estimation as follows: if the maximal standardized residual is tested as mass or obvious suspicious mass, the bifactor equivalent weight is only constructed for the observation with mass, while the weight of other observations remains unchanged. In addition to the two cases above, the bifactor equivalent weight can be constructed for all observations at the same time. The calculation steps can also be simplified based on the improved equivalent weight construction method. Simulation calculation and analysis are carried out using the observation data of a GPS network. The results demonstrate that the method can improve calculation efficiency when containing a small amount of masses, and improve the accuracy of parameter estimation when containing more masses.
