Abstract
Our proposed method introduces a hierarchical Bayesian regression model with an effective Gibbs sampler for parameter estimation and variable selection. It combines a regularization technique with Bayesian inference by reformulating the Laplace distribution as a mixture of Uniform and Exponential distributions. This mixed distribution is incorporated into the composite quantile regression model to form a new hierarchical model with strong statistical properties.
Through a simulation study, our method outperformed both Bayesian and non-Bayesian approaches in variable selection and estimation. Additionally, it demonstrated excellent stability. When applied to real-world Thrombocytopenia data, our method proved highly effective in both estimating parameters and selecting relevant variables.
Through a simulation study, our method outperformed both Bayesian and non-Bayesian approaches in variable selection and estimation. Additionally, it demonstrated excellent stability. When applied to real-world Thrombocytopenia data, our method proved highly effective in both estimating parameters and selecting relevant variables.