生存资料回归模型分析——生存资料及其统计分析方法概述  

Analysis of regression model of survival data——overview of survival data and its statistical analysis methods

作　　者：刘媛媛[1] 李长平[1,2] 胡良平 Liu Yuanyuan;Li Changping;Hu Liangping(School of Public Health,Tianjin Medical University,Tianjin 300070,China;Specialty Committee of Clinical Scientific Research Statistics of World Federation of Chinese Medicine Societies,Beijing 100029,China;Graduate School,Academy of Military Sciences PLA China,Beijing 100850,China)

机构地区：[1]天津医科大学公共卫生学院,天津300070 [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 [3]军事科学院研究生院,北京100850

出　　处：《四川精神卫生》2020年第1期21-26,共6页Sichuan Mental Health

基　　金：国家自然科学基金项目(项目名称：贝叶斯生存分析方法在肝细胞癌肝移植患者预后预测中的应用研究，项目编号：81803333)

摘　　要：本文目的是全面介绍生存资料的特点及其常用统计分析方法。生存资料具有以下四个特点:①同时具有生存结局和生存时间;②生存时间可能含有删失数据或截尾数据;③生存时间的分布通常不服从正态分布,常呈指数分布、Weibull分布、对数正态分布;④影响生存时间的因素较复杂且不易控制。生存资料统计分析方法涉及统计描述、差异性分析和回归分析三大类,其中,统计描述主要有Kaplan-Meier(卡普兰-迈耶)估计法和Life table(寿命表)估计法;差异性分析主要有对数秩检验(log-rank test)和威尔考克森检验(Wilcoxon test);而回归分析主要有Cox比例和非比例风险回归模型、参数回归模型。在对生存资料进行统计分析时,需要合理选择统计分析方法,方可全面而又深入地揭示生存资料的内在变化规律。The purpose of this paper was to provide a comprehensive introduction to the characteristics of survival data and its commonly used statistical analysis methods.Survival data had the following four prominent features:①survival data had both survival outcomes and survival times;②the survival times might contain censored data or truncated data;③the distribution of survival times usually did not follow the normal distribution,and the survival times were often distributed in the exponential distribution,Weibull distribution,lognormal distribution,and so on;④the factors affecting the survival times were complicated and difficult to control.The statistical analysis methods of survival data involves three categories:statistical description,difference analysis and regression analysis.The statistical description methods mainly included Kaplan-Meier estimation method and Life table estimation method.The difference analyses mainly included log-rank test and Wilcoxon test.The regression analyses mainly included Cox’s proportional and non-proportional hazards regression models and parametric regression models.In the statistical analysis of the survival data,it was necessary to select the statistical analysis methods reasonably,in order to comprehensively and deeply reveal the inherent changes of the survival data.

关 键 词：生存资料 删失数据 生存分析 生存函数 风险函数 回归分析 

分 类 号：R195.1[医药卫生—卫生统计学]