Fully Sampling-Based Bayesian Elastic Network Linear Quantitle Regression Models

We propose a fully sampling-based Bayesian method to analyze the elastic net quantile regression models.As the full conditional posterior of the regularized coefficients contains an intractable factor,the existing method approximates it by means of the numerical method,which not only is time-consuming,but also leads to the bias of the approximation.We develop an exchange algorithm to address these problem.Moreover,we make use of the partially-collapsed technique to speed up the convergence of our algorithm.Simulation studies verify the efficiency and practicability of the proposed approach.We apply the proposed method to a real data set.