基于逆跳MCMC的贝叶斯分位自回归模型研究

Bayesian Quantile Autoregressive Models Using Reversible Jump MCMC Algorithm

考虑到传统信息理论方法确定模型存在不足,在贝叶斯理论框架下提出了基于逆跳马尔可夫链蒙特卡罗法确定分位自回归模型阶次的方法。在时间序列服从非对称Laplace分布的条件下,设计了马尔可夫链蒙特卡罗数值计算程序,得到了不同分位数下模型参数的贝叶斯估计值。实证研究表明:基于逆跳马尔可夫链蒙特卡罗法的贝叶斯分位自回归模型能有效地揭示滞后变量对响应变量的位置、尺度和形状的影响。

With the deficiency of traditional modeling method based on information theory,this paper gives a quantile autoregressive model based on Reversible Jump Markov Chain Monte Carlo in the theoretical framework of Bayesian.Supposing that time series subject to asymmetric Laplace distribution,Markov chain Monte Carlo numerical simulation program was designed,and the quantile autoregressive parameters were estimated.Empirical studies have shown that Bayesian quantile autoregression based on Reversible Jump Markov Chain Monte Carlo can effectively reveal the lagged variables effect the location,scale and shape of the response variable.