基于Gibbs抽样算法的面板数据分位回归方法

Quantile Regression for Panel Data Based on Gibbs Sampling Algorithm

文章讨论了含有随机效应的面板数据模型,利用非对称Laplace分布与分位回归之间的关系,文章建立了一种贝叶斯分层分位回归模型。通过对非对称Laplace分布的分解,文章给出了Gibbs抽样算法下模型参数的点估计及区间估计,模拟结果显示,在处理含随机效应的面板数据模型中,特别是在误差非正态的情况下,本文的方法优于传统的均值模型方法。文章最后利用新方法对我国各地区经济与就业面板数据进行了实证研究,得到了有利于宏观调控的有用信息。

The paper discusses the random effects panel data model and establishes a hierarchical Bayesian quantile regression model by using of the relationship between asymmetric Laplace distribution （ALD） and quantile regression. The point and interval estimate of unknown parameters are obtained by Gibbs sampling algorithm based on decomposition of ALD. Monte Carlo simulation study also indicates that the proposed method is better than mean regression methods when dealing with the panel data model with random effects, especially when the error term is non - normal. Finally, we use the new method to study an economy and employment panel data of our country and obtain much useful information for macroeconomic control.