现状数据下的贝叶斯比例风险模型的变量选择

Variable Selection of Bayesian Proportional Hazards Model Based on Current Status Data

本文在贝叶斯框架下考虑现状数据比例风险模型的变量选择问题。首先构造基于spikeand slab先验,运用二元潜变量标记活跃协变量,给出满条件分布及相应的Gibbs抽样算法。数值模拟比较了该方法与Lasso、SCAD和ALasso方法,结果表明该方法模型正确识别率高。实例选用Ⅱ型糖尿病患者心脏衰竭数据,分析选择出最显著的影响因素,验证了该方法的有效性。

Inthis paper,we consider the variable selection problem of current status data proportional hazards model in the framework of Bayesian Firstly based on spike and slab priors,the active covariates are marked by binary latent variables,and the full conditional distribution and the corresponding Gibbs sampling algorithm are given.The simulation study results show that the proposed method has high recognition rate compared with Lasso,SCAD and Adaptive Lasso methods.The real data of heart failure disease for Type II diabetic patients were used to analyze and select the most significant influencing factors,which verified the effective ness of the method.