No Cover Image

Journal article 1041 views

A Simple Diagnostic Method of Outlier Detection for Stationary Gaussian Time Series

Yuzhi Cai Orcid Logo, Davies, Neville

Journal of Applied Statistics, Volume: 30, Pages: 205 - 223

Swansea University Author: Yuzhi Cai Orcid Logo

Abstract

In this paper we present a ``model free'' method of outlier detectionfor Gaussian time series by using the autocorrelation structure of thetime series. We also present a graphic diagnostic methodin order to distinguish an AO from IO. The test statistic fordetecting the outlierhas a $\chi^2...

Full description

Published in: Journal of Applied Statistics
Published: 2003
URI: https://cronfa.swan.ac.uk/Record/cronfa15298
Tags: Add Tag
No Tags, Be the first to tag this record!
Abstract: In this paper we present a ``model free'' method of outlier detectionfor Gaussian time series by using the autocorrelation structure of thetime series. We also present a graphic diagnostic methodin order to distinguish an AO from IO. The test statistic fordetecting the outlierhas a $\chi^2$ distribution with one degree of freedom. We show thatthis method works well when time series contain either one type of theoutliers or both additive andinnovation type outliers, and this method has the advantage that notime seriesmodel needs to be estimated from the data. Simulation evidence showsthat different types of outliers can be graphically distinguishedby using the techniques proposed.
Keywords: diagnostic method, additive outliers, innovation outliers, simulation, Gaussian time series.
College: Faculty of Humanities and Social Sciences
Start Page: 205
End Page: 223