1: \begin{abstract}

2: Quantile regression is

3: often used when a comprehensive relationship between a

4: response variable and one or more explanatory variables is desired. The

5: traditional frequentists' approach to quantile regression has been

6: well developed around asymptotic theories and efficient

7: algorithms. However, not much work has been published under the

8: Bayesian framework. One challenging problem for Bayesian

9: quantile regression is that the full likelihood has no parametric forms.

10: In this paper, we propose a Bayesian quantile

11: regression method, the linearly interpolated density (LID) method,

12: which uses a linear interpolation of the quantiles to approximate the

13: likelihood.

14: Unlike most of the existing methods that aim at tackling one

15: quantile at a time, our proposed method estimates the

16: joint posterior distribution of multiple quantiles, leading to

17: higher global efficiency for all quantiles of interest.

18: Markov chain Monte Carlo algorithms are developed to carry out

19: the proposed method.

20: We provide convergence results that justify both the algorithmic convergence

21: and statistical approximations to an integrated-likelihood-based posterior.

22: From the simulation results, we verify that LID

23: has a clear advantage over other existing methods in estimating

24: quantities that relate to two or more quantiles.

25: \end{abstract}