基于GED的混合ARCH模型  

A Mixture Autoregressive Conditional Heteroscedastic Model with Generalized Error Distribution

作　　者：田晓兰[1] 柳金甫[2] 

机构地区：[1]安徽财经大学统计与应用数学学院,蚌埠市233030 [2]北京交通大学理学院,北京100044

出　　处：《北京交通大学学报》2007年第3期87-90,共4页JOURNAL OF BEIJING JIAOTONG UNIVERSITY

摘　　要：提出了基于广义误差分布的混合自回归条件异“方差”模型,将时间序列尾部的特征融入到混合条件异“方差”模型的参数估计之中,发现模型中的指标r和数据本身尾部厚薄的性质有关.给出了该模型参数估计的EM算法,并利用BIC准则对模型进行定阶.A new mixture autoregressive conditional heteroscedastic model with a generalized error distribution, is proposed for modelling nonlinear time series. The estimation of the parameters of the mixture autoregressive conditional heteroscedastic model takes the time series heavy-tailed nature into account, and detect that the index r corresponding to the tail-thickness index. The estimation of the parameters can be easily done through EM algorithm and the order model is also easily selected by BIC criterion.

关 键 词：广义误差分布 混合ARCH模型 EM算法 厚尾性 

分 类 号：O241.84[理学—计算数学]