用SAS软件实现高维列联表资料的统计学分析:因变量为二值变量的多重logistic回归分析  

Statistical analysis and SAS solutions for multi-dimensional contingency table:multiple logistic regression analysis of binary dependent varibles

作　　者：胡良平[1] 沈宁[1] 柳伟伟[1] 

机构地区：[1]军事医学科学院研究生部生物医学统计学咨询中心,北京100850

出　　处：《药学服务与研究》2014年第1期6-10,共5页Pharmaceutical Care and Research

基　　金：重大新药创制国家科技重大专项课题资助项目(2011ZX09302-006-01)

摘　　要：Logistic回归分析属于概率型回归分析,适用于因变量为定性变量的数据分析和建模,但对自变量的数目和性质没有特殊要求。因变量为二值变量的多重logistic回归分析适用于因变量编码为0或1(代表阳性或者阴性)的多重logistic回归分析。从整体上理解logistic回归分析,可根据操作过程依次总结为以下几个方面:自变量筛选、建立回归模型、进行假设检验(包括对回归系数的检验、整体模型检验以及模型拟合优度检验)。近年来,logistic回归分析在众多临床医学研究,尤其是在预测因子分析中得到了广泛应用,但存在一些问题,例如缺少对模型拟合优度检验以及后续通过验证集对其预测效能进行的二次检验。本文重点介绍如何正确实施多重logistic回归分析、其SAS实现及结果解读。Being able to model the probability of success or failure, the logistic regression analysis requires the dependent variable to be qualitative in nature,while it has no specific requirements for the number and nature of the independent variables. As a type of the multiple logistic regression analysis, binary multiple logistic regression analysis is suitable for the dependent variables coded by 0 or 1, which represents positive or negative results, respectively. In order to understand the logistic regression analysis in a rapid way, the process of logistic regression analysis is summarized in the following steps：screening the dependent variables,establishing regression model,and hypothesis tests including the tests of regression coefficients, the entire regression model and goodness of fit. Although logistic regression analysis has been widely used in clinical medical researches in recent years, especially in predicting risk factors, problems still exist, including lack of evaluating goodness of fit and following re-evaluation of prediction efficiency by a validation set. This paper focuses on how to carry out a muhiple logistic regression analysis with SAS software in a proper way, and how to explain the results in combination with clinical work.

关 键 词：统计学 LOGISTIC回归分析 二值变量 SAS实现 

分 类 号：O212.6[理学—概率论与数理统计] R195.1[理学—数学]