分段指数模型的拟合及SAS实现

The Fitting of the Piecewise Exponential Model and its SAS Procedures

目的以实例阐述生存分析分段指数模型的拟合及SAS实现。方法将生存时间划分为几个区间,利用SAS中的PROCGENMOD过程或PROCLIFEREG过程,对生存资料进行分段指数模型拟合,并通过参数估计结果,计算各时间区间的风险率及生存率。结果该资料拟合分段指数分布的结果比较满意,三个时间区间的死亡风险均不相同,且第一个和第二、三个区间的差别有统计学意义(P<0.05)。结论分段指数模型不仅模型形式简单,参数易于估计,若拟合得当可提高统计效率。

Objective To introduce the fitting of piecewise exponential model in survival analysis and its corresponding SAS procedures. Methods Divide the survival time scales into severn intervals first. Then by PROC GENMOD procedure or PROC LIFEREG procedure in SAS, piecewise exponential model can be built for the survival data. Hazard function and survival function of each time interval were calculated based on the results of parameter estimation. Results The result of fitting was relative satisfactory. The hazards of death in three intervals were different. Differences between the fast and the second, the fast and the third interval were statistically significant. Conclusion Not only is the form of the piecewise exponential model simpler, but also it＇ s easier to estimate parameters. If the model is appropriately fitted,the statistical efficiency can be improved.