双自回归模型下基于EM算法的分位数回归分析

Quantile regression analysis for double autoregressive model using EM algorithm

在似然框架下研究双AR(p)分位数回归的统计推断,基于正态指数混合方法修正误差分布,构造EM算法求解此模型的参数估计,通过BIC准则定阶,考察其在不同分位点上的表现。模拟研究表明,所提EM算法在双AR(p)分位数回归估计中表现出色,BIC定阶较为准确。最后对沪深300指数收盘价进行股价分析和预测。

The statistical inference of double AR(p)quantile regression was studied under the likelihood framework,and the error distribution was modified based on the normal and exponential mixed method.The EM algorithm was constructed to estimate the parameter of the model.The performance of the model at different quantile was investigated by the BIC criterion.Simulation studies show that the proposed EM algorithm performs well in the double AR(p)quantile regression estimation,and the degree of BIC criterion is relatively accurate.Finally,the model is applied to analyze and forecast the stock price of the CSI 300 Index.