随机删失下半参数模型的补偿估计方法  

Penalized least squares for semiparametric models under stochastic censored

作　　者：潘雄[1] 孙海燕[1] 

机构地区：[1]武汉大学测绘学院

出　　处：《测绘科学》2005年第4期27-29,共3页Science of Surveying and Mapping

基　　金：国家自然科学基金资助项目(40274005);湖北省自然科学基金资助项目(2004ABA032)

摘　　要：考虑半参数测量模型L=Bx+S+Δ,x∈Rd为未知回归参数,S为未知Borel函数。本文首先利用自然样条函数法,找到符合条件的非参数自然插值样条函数。其次利用补偿法并综合最小二乘法,导出了这种平差方法的解算公式。在本文的最后,将这种方法与最小二乘平差方法进行了比较分析,结果说明,半参数测量模型能更接近于真实情况。In this paper,based on penalized least squares,the penalized weighted sum of squares is set up;we deduce the Semi-parametric adjustment models under censored.Natural cubic spline is used to find nonparametric functions.The estimators are derived by using the penalized least squares.According to the properties of the random errors,we study the choice of a reasonable smoothing parameter depending on the mixed model of penalized splines.At the end,a simulation is used to testify the method.It is clearly shown that semiparametric adjustment with a suitable parameter is better that the least squares.The adaptation of this technique is more flexible than the traditional model.

关 键 词：半参数模型 随机删失 补偿最小二乘估计法 

分 类 号：O212.7[理学—概率论与数理统计] P207[理学—数学]