一种建立回归模型的新方法  

A NEW METHOD FOR ESTABLISHING REGRESSION MODEL

作　　者：许双安[1] 姚宜斌[1] 孔建[1] 龚佩佩[1] 

机构地区：[1]武汉大学测绘学院,武汉430079

出　　处：《大地测量与地球动力学》2010年第4期117-121,共5页Journal of Geodesy and Geodynamics

基　　金：国家自然科学基金(40774008);国家973计划项目(2006CB701301)

摘　　要：基于经典最小二乘方法在实际应用中存在的局限性,提出一种建立回归模型的新方法,即先利用传统方法建立近似回归方程,然后再用整体最小二乘方法对近似回归方程进行修正。仿真数据和变形监测实测数据验算证明,该方法能有效地提高形变预报精度。The regression method is used to study the related but not determined relations between variables,which is widely used in the field of deformation monitoring data processing and analyzing.The method based on classical least-squares under the indirect adjustment model or on singular value decomposition under the overall least-squares criterion for solving vegression parameters has some limitations in practical applications,it may weaken the statistical significance of the regression model and the significance test even will fail.A new method is proposed that is to establish the approximate regression equation by the traditional method first,then amendment the regression model by the Total Least-Squares.The feasibility and correctness of the new method are verified by the simulation data and measured real deformation data and the prediction accuracy can be improved.

关 键 词：回归模型 经典最小二乘 整体最小二乘 显著性检验 预报分析 

分 类 号：P207[天文地球—测绘科学与技术]