基于MCMC的分位回归GARCH模型的贝叶斯分析

Bayesian Analysis of Quantile GARCH Models Based on MCMC Algorithm

基于最小一乘方法提出了一类分位回归GARCH模型的2步估计方法,并且基于双指数分布和非对称Laplace分布构建了GARCH模型的似然函数,选择扩散先验分布,实现了对模型的贝叶斯估计.仿真分析发现基于最小一乘方法的贝叶斯分位回归方法可以全面有效地实现对GARCH模型的估计.

In the report,based on least absolute deviation,the two step estimation method for GARCH models using quantile regression was proposed,and based on double exponential distribution and asymmetric Laplace distribution,the likelihood function of GARCH model was constructed,dispersion and prior distribution was selected,and Bayes estimation for model was achieved. The simulation results showed that Bayes quantile regression estimation is effective to achieve the estimation for GARCH model.