零膨胀Poisson分布模型回归分析  

The regression analysis of the zero-inflated Poisson distribution model

作　　者：胡良平 Hu Liangping(Graduate School,Academy of Military Sciences PLA China,Beijing 100850,China;Specialty Committee of Clinical Scientific Research Statistics of World Federation of Chinese Medicine Societies,Beijing 100029,China)

机构地区：[1]军事科学院研究生院,北京100850 [2]世界中医药学会联合会临床科研统计学专业委员会,北京100029

出　　处：《四川精神卫生》2018年第5期405-410,共6页Sichuan Mental Health

基　　金：国家高技术研究发展计划课题资助(2015AA020102)

摘　　要：本文目的是介绍零膨胀Poisson分布模型回归分析。首先,介绍零膨胀计数资料及其零膨胀Poisson分布回归模型构建原理,包括"零膨胀Poisson分布回归模型的形式"和"零膨胀Poisson分布回归模型的求解";其次,介绍"零膨胀Poisson分布回归模型的SAS实现",包括"创建SAS数据集""呈现因变量Y的频数分布""求出因变量Y的均值和方差"和"基于全部自变量对因变量Y构建多重零膨胀Poisson分布回归模型"。本文结果提示,当计数资料为非严重过离散的零膨胀计数资料时,拟合"多重零膨胀Poisson分布回归模型",可获得满意的拟合效果。The purpose of this paper was to introduce the regression analysis of the zero-inflated Poisson distribution model.Firstly,the concepts of the zero-inflated count data and the building principle of the zero-inflated Poisson distribution regression model were given,which included the following two aspects：（1）the form of the zero-inflated Poisson distribution regression model;（2）the solution for the model mentioned before. Secondly,the SAS realization of this kind of model for the zero-inflated count data was presented. The contents were as follows：（1）creating SAS data set;（2）displaying the frequency of the count dependent variable Y;（3）calculating the arithmetic mean and variance of the count dependent variable Y;（4） building a multiple zero-inflated Poisson distribution regression model based on all independent variables. The results of the article showed that the satisfaction of the fitted effects could be gotten by building the zero-inflated Poisson distribution regression model to the zero-inflated count data with the not severe over-dispersion.

关 键 词：零膨胀计数资料 过离散 零膨胀Poisson分布回归模型 概率函数 极大似然估计 

分 类 号：R195.1[医药卫生—卫生统计学]