Logistic回归模型的经验似然诊断  

Statistic Measures for Logistic Regression Models Based on Empirical Likelihood Method

作　　者：丁先文[1,2] 陈绚青[1] 周友明[1] 邹舒[1] 

机构地区：[1]江苏理工学院数理学院,江苏常州213001 [2]东南大学数学系,江苏南京211189

出　　处：《数学的实践与认识》2013年第9期140-148,共9页Mathematics in Practice and Theory

基　　金：江苏理工学院博士启动基金(KYY11052)

摘　　要：经验似然方法已经被广泛用于许多模型的统计推断.基于经验似然对Logistic回归模型进行统计诊断.首先给出模型的估计方程,进而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;最后通过实例分析,说明了统计诊断方法的有效性.The empirical likelihood method has been extensively applied to many models of statistical inference. This paper is based on empirical likelihood for Logistic Regression models for statistical diagnosis. First, the estimating equations of the model are given and the maximum empirical likelihood estimates of the parameters are obtained; then, based on empirical likelihood method, the three different measures of influence curvatures are studied; last, data analysis are given to illustrate the validity of statistical diagnostic measures.

关 键 词：经验似然 估计方程 极大经验似然估计 影响曲率 

分 类 号：O212.1[理学—概率论与数理统计]