期刊
COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION
卷 46, 期 8, 页码 5861-5878出版社
TAYLOR & FRANCIS INC
DOI: 10.1080/03610918.2016.1183780
关键词
Asymmetric Laplace distribution; EM algorithm; Gibbs sampling; Inverse Bayes formulae; Quantile regression
资金
- National Science Foundation of China [11371227]
In this article, a non-iterative posterior sampling algorithm for linear quantile regression model based on the asymmetric Laplace distribution is proposed. The algorithm combines the inverse Bayes formulae, sampling/importance resampling, and the expectation maximization algorithm to obtain independently and identically distributed samples approximately from the observed posterior distribution, which eliminates the convergence problems in the iterative Gibbs sampling and overcomes the difficulty in evaluating the standard deviance in the EM algorithm. The numeric results in simulations and application to the classical Engel data show that the non-iterative sampling algorithm is more effective than the Gibbs sampling and EM algorithm.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据