基于删失中位数回归的贝叶斯经验似然

Bayesian empirical likelihood for median regression model with censored data

将经验似然纳入贝叶斯框架,在删失中位数回归模型下提出了模型参数的贝叶斯经验似然估计,证明了模型参数的后验分布是渐近正态的.运用M-H算法得到了点估计以及置信域,可以避免直接优化经验似然的繁重任务.模拟研究比较了贝叶斯经验似然与LAD和经验似然在有限样本下的表现,结果展示出贝叶斯经验似然优于LAD估计和经验似然.

By taking the empirical likelihood into a Bayesian framework,the Bayesian empirical likelihood estimation for the parameters of the censored median regression model is proposed.It is shown that the resultant posterior is asymptotically normal.By using the MH algorithm to obtain the estimator and the confidence region,the daunting task of directly maximizing empirical likelihood is avoided.The simulation experiments were conducted to compare the Bayesian empirical likelihood and the LAD and empirical likelihood under finite samples.The simulation results demonstrate that Bayesian empirical likelihood performs better than the LAD and empirical likelihood in practical situations.