ARMAX时间序列模型异常点及异常点斑片的估计和检测

将通常的Gibbs抽样和自适应的Gibbs抽样算法用于带有外生变量的自回归移动平均时间序列(ARMAX)模型的Bayes分析,首先采用一些方法消除ARMAX模型中输入(外生变量)序列的影响,然后在前人工作的基础上给出了一种类似的挖掘相应时间序列中的异常点及异常点斑片的方法.说明了自适应的Gibbs抽样算法也能够有效地检测ARMAX模型中孤立的附加型异常点及异常点斑片.实际的和模拟的结果也显示这些方法可以明显减少掩盖和淹没现象的发生,这是对已有工作的推广和扩充.

ESTIMATION AND DETECTION OF OUTLIERS AND PATCHES IN ARMAX TIME SERIES MODELS

In this paper, the usual Gibbs sampler and adaptive Gibbs sampler are used in the Bayesian analysis of autoregressive moving average with exogenous variable(ARMAX) time series models. Firstly, some methods are used to delete theinfluence of input process in ARMAX model, and then the method of mining outliersand patches in time series based on the former work is given. It is shown that the adaptive Gibbs sampler is also useful in handling additive isolated outliers and outlier patches in ARMAX model. Practical and simulation studies also show that the procedure can reduce possible masking and swamping effects, and hence improve the existing methods.