关于Logistic回归中最大似然估计的可估性  

ESTIMATION OF MAXIMUM LIKELIHOOD IN LOGISTIC REGRESSION MODEL

作　　者：张彦荣[1] 王彤[1] 

机构地区：[1]山西医科大学卫生统计教研室,太原030001

出　　处：《现代预防医学》2009年第10期1807-1809,共3页Modern Preventive Medicine

基　　金：教育部科学技术研究重点项目(206020);山西省高校青年学术带头人基金(晋教科2004-13号)

摘　　要：[目的]分析Logistic回归中最大似然估计的可估性与样本点布局的关系。[方法]结合样本空间数据点布局的几何表现,阐述分离和有重叠资料Logistic模型参数最大似然估计的可估性。[结论]Logistic回归中最大似然估计的可估性与观察空间各样本点的布局有关,结合统计软件提示信息判断样本点的布局是考察似然估计是否存在的重要步骤。[Objective] To explore the relationship between the existence of maximum likelihood estimation and the configurations of the sample space in Logistic regression model. [Methods] We describe the estimation of maximum Likelihood for separation and overlapping data with the geometric presentation of data arrangement of sample space. [ Conclusion ] The estimation of maximum Likelihood in Logistic regression medel has association with the configurations of the sample space. Therefore, it is an important step to detect the configurations of the sample points with the aids of the warning messages from software outputs.

关 键 词：LOGISTIC回归 最大似然估计 完全分离 拟完全分离 重叠 

分 类 号：R195.1[医药卫生—卫生统计学]