非线性时间序列建模的均值异方差混合转移分布模型

进一步研究了了由Berchtold提出的均值异方差混合转移分布(expectation heteroscedastic mixture transition distribution model,EHMTD)模型.讨论并得到了EHMTD模型的平稳性条什和分布函数的尾部特征.运用ECM(expectation conditional maximization)算法估计模型的参数.条件分布的多样性使得该类模型能够对非对称、多峰、厚尾等非Gauss特征进行描述.模拟及实例分析的结果表明EHMTD模型是一类易于建模,并且有着广泛应用前景的非线性时间序列模型.

Expectation heteroscedastic mixture transition distribution model for modeling nonlinear time series

The expectation heteroscedastic mixture transition distribution (EHMTD) model first introduced by Berchtold is further studied in this paper.First,the stationary conditions and tail behaviors of the model are derived.The estimation of parameters is easily performed via expectation conditional maximization (ECM) algorithm.The variety of conditional distributions of the EHMTI) model makes the model capable of modeling time series with asymmetric multimodal or heavier tail distribution.The model is applied to simulate real data sets with satisfactory results.The EHMTD model is easy to model and potentially useful in modeling nonlinear time series.