非线性回归模型的经验似然诊断  

Diagnostic Measures for Nonlinear Regression Models Based on Empirical Likelihood Method

作　　者：丁先文[1] 徐亮[2] 林金官[2] 

机构地区：[1]江苏技术师范学院数理学院,常州213001 [2]东南大学数学系,南京211189

出　　处：《应用数学学报》2012年第4期693-702,共10页Acta Mathematicae Applicatae Sinica

基　　金：国家自然科学基金(11171065;11102076);江苏省自然科学基金(BK2011058);江苏技术师范学院青年科研基金(KYY11073)资助项目

摘　　要：经验似然方法已经被广泛用于线性模型和广义线性模型.本文基于经验似然方法对非线性回归模型进行统计诊断.首先得到模型参数的极大经验似然估计;其次基于经验似然研究了三种不同的影响曲率度量;最后通过一个实际例子,说明了诊断方法的有效性.The empirical likelihood method has been extensively applied to linear regres- sion and generalized linear regression models. In this paper, the diagnostic measures for nonlinear regression models are studied based on the empirical likelihood method. First, the maximum empirical likelihood estimate of the parameters are obtained. Then, three different measures of influence curvatures are studied. Last, real data analysis are given to illustrate the validity of statistical diagnostic measures.

关 键 词：经验似然 极大经验似然估计 影响曲率 

分 类 号：O212.2[理学—概率论与数理统计]