【摘要】 针对当前状态数据中的比例风险模型提出了一种基于期望-最大化的贝叶斯变量选择方法.该模型能够同时进行参数估计和变量选择,有效地增强了模型的可解释性和预测能力.为了识别风险因素,首先对表示协变量是否存在的指示变量赋予合适的先验分布,使用单调样条来近似基准累积风险函数;然后通过使用基于泊松隐变量的两阶段数据扩充技术提出了一种有效的期望-最大化对模型拟合算法;最后通过模拟研究和一个实例分析证明了所提方法的有效性.更多还原

【Abstract】 A Bayesian proportional hazards(PH) model is proposed for analyzing current status data based on Expectation-Maximization Variable Selection(EMVS) method. This model can estimate parameters and select variables simultaneously, which efficiently improves model interpretability and predictive ability. To identify risk factors, appropriate priors are assigned on the indicator variables that denote the existence of covariates. The baseline cumulative hazard function is approximated via monotone splines. A novel Expectation-Maximization(EM) algorithm is developed for model fitting by using a two-stage data augmentation procedure involving latent Poisson variables. Finally, the performance of proposed method is investigated by simulations and a real data analysis.更多还原