半参数模型稳健估计及其应用  

Robust Estimate and Its Applications for Semiparametric Models

作　　者：丁士俊[1] 张松林[2] 张洪波[1] 陶本藻[1] 

机构地区：[1]武汉大学测绘学院,湖北武汉430079 [2]同济大学测量与国土信息系,上海200092

出　　处：《测绘科学技术学报》2007年第1期26-29,共4页Journal of Geomatics Science and Technology

基　　金：国家自然科学基金资助项目(40604001)

摘　　要：在半参数模型估计中,均假设观测误差服从正态分布。当观测量含有粗差时,粗差对参数和非参数估计的影响是不可忽略。基于此,首先在总结线性参数模型稳健估计基本理论的基础上,论述了M估计权因子的确定方法。然后提出了半参数模型稳健估计方法,并导出半参数模型(广义)补偿最小二乘稳健估计的基本公式。最后通过两个模拟算例验证了其估计方法的有效性。In the estimation of semiparametric models, observation errors are assumed to be subject to normal distribution. When observations contain gross errors, affection on the estimates of parameter and nonparameter for gross errors can be not ignored. Summing up the basic theories of the robust estimates for linear parametric models, this paper discussed the determination method of the weight-factor for M estimates, put forward the robust estimates of the semiparametric regression model, and derived the basic formulae of the robust estimates for the semiparametric models. Finally with two simulation examples the validity of the estimations were verified.

关 键 词：半参数模型 补偿最小二乘 广义补偿最小二乘 稳健估计 

分 类 号：P207[天文地球—测绘科学与技术]