Asymptotic estimation theory of change-point problems for time series regression models and its applications
Asymptotic estimation theory of change-point problems for time series regression models and its applications
复制标题
时间序列回归模型变点问题的渐近估计理论及其应用
DOI：
10.1214/lnms/1215091669
复制
发表时间：
2003
期刊：
影响因子：
--
通讯作者：
M. Puri
中科院分区：
文献类型：
--
作者：
Takayuki Shiohama;M. Taniguchi;M. Puri
在时间序列模型的趋势中，如何发现结构变化是一个重要的问题.本文研究了具有圆形阿尔马残差的时间序列回归模型趋势变点的估计问题。首先，我们展示了连续假设之间似然比的渐进性。其次，我们构造了包含变点的未知参数的极大似然估计（MLE）和贝叶斯估计（BE）。然后，它表明，建议的BE是渐近有效的，而MLE是不那么普遍。文中还给出了数值研究和应用。AMS科目分类：62M10、62M15、62N99
It is important to detect the structural change in the trend of time series model. This paper addresses the problem of estimating change point in the trend of time series regression models with circular ARMA residuals. First we show the asymptotics of the likelihood ratio between contiguous hypotheses. Next we construct the maximum likelihood estimator (MLE) and Bayes estimator (BE) for unknown parameters including change point. Then it is shown that the proposed BE is asymptotically efficient, and that MLE is not so generally. Numerical studies and the applications are also given. AMS subject classifications: 62M10, 62M15, 62N99