生存分析中的解释变异度量及其应用  

Explained Variation and its Application in Survival analysis

作　　者：余红梅[1] 何大卫[1] 

机构地区：[1]山西医科大学卫生统计学教研室,030001

出　　处：《中国卫生统计》2002年第5期258-260,共3页Chinese Journal of Health Statistics

基　　金：全国统计科学研究项目 (990 1 9) ;山西省归国人员基金项目

摘　　要：目的　介绍生存分析中解释变异度量及其应用。方法　应用平方和定义的R2 、模型似然定义的R2 及基于预测值和生存的平方相关的R2 度量解释变异大小。结果　上述三类不同的R2 可度量生存函数的变异中利用协变量可减少的百分比。结论　没有给定预后因子下解释变异的量化 ,研究者很容易被高度显著的P值或预后因子的高相对危险度所误导 ,过高地估计一个预后模型的预测能力。Objective To introduce explained variation and its application in survival analysis.Methods We use the sums-of-squares defintion of R \+2?definition of R \+2 by model likelihoods and R\+2 measures based on the squared correlation of the predictor and of survival to measure explained variation.Results The above three types of R \+2 can measure how much of the variation in S(t) is reduced by using the covariates.Conclusion Without quantification of the variation explained by a given set of prognostic factors,investigators are easily misled by highly significant P -values or impressive relative risk estimates for prognostic factors and assess highly the prognostic ability of a model.

关 键 词：生存分析 COX模型 解释变异 预测 量化 

分 类 号：R195[医药卫生—卫生统计学]