您的位置:首页  > 论文页面

运用极值理论挖掘时间序列模型中的异常点

发表时间:2008-03-15  浏览量:2363  下载量:836
全部作者: 田玉柱,陈平
作者单位: 天水师范学院数学与统计学院;东南大学数学系
摘 要: 用检验的方法诊断时间序列异常点的关键就是决定检验统计量在一定的显著性水平下是否超越某一临界值,临界值的选取一般文献都描述的很模糊。本文运用极值理论给出诊断AR(p)模型异常点选取临界值的分布近似方法,这种方法选取的临界值可保证控制在一定的显著性水平下,而且还可以计算出检验的渐近p值,比一般仿真选取的临界值更科学合理。
关 键 词: 概率论与数理统计;时间序列分析;极值分布;AR(p)模型;IO异常点;AO异常点
Title: Outlier detection in time series based on extreme value theory
Author: TIAN Yuzhu, CHEN Ping
Organization: Department of Mathematics and Statistics, Tianshui Normal College;Department of Mathematics, Southeast University
Abstract: The detection of outliers in time series hinges on determining whether the test statistic exceeds a critical value for a given significance level. This value is prescribed only vaguely by many authors with no reference to any significance level. This paper will give out an asymptotic critical value in a fixed significance level and an asymptotic p-value for testing by means of extreme value theory in order to detect outliers in AR(p) model. This method is better than the simulation and more scientific.
Key words: probability and mathematical statistics; time series analysis; Extreme value distribution; AR(p)model; IO outlier; AO outlier
发表期数: 2008年7月第5期
引用格式: 田玉柱,陈平. 运用极值理论挖掘时间序列模型中的异常点[J]. 中国科技论文在线精品论文,2008,1(5):575-581.
 
5 评论数 0
暂无评论
友情链接