Additive Outlier Detection via Extreme-Value Theory.

Peter Burridge; A. M. Robert Taylor

doi:10.1111/j.1467-9892.2006.00483.x

Additive Outlier Detection via Extreme-Value Theory.

Peter Burridge, A. M. Robert Taylor

Economics

Research output: Contribution to journal › Article › peer-review

Abstract

This article is concerned with detecting additive outliers using extreme value methods. The test recently proposed for use with possibly non-stationary time series by Perron and Rodriguez [Journal of Time Series Analysis (2003) vol. 24, pp. 193–220], is, as they point out, extremely sensitive to departures from their assumption of Gaussianity, even asymptotically. As an alternative, we investigate the robustness to distributional form of a test based on weighted spacings of the sample order statistics. Difficulties arising from uncertainty about the number of potential outliers are discussed, and a simple algorithm requiring minimal distributional assumptions is proposed and its performance evaluated. The new algorithm has dramatically lower level-inflation in face of departures from Gaussianity than the Perron–Rodriguez test, yet retains good power in the presence of outliers.

Original language	English
Pages (from-to)	685-701
Number of pages	17
Journal	Journal of Time Series Analysis
Volume	27
Issue number	5
DOIs	https://doi.org/10.1111/j.1467-9892.2006.00483.x
Publication status	Published - 21 Apr 2006

Keywords

Additive outliers • extreme order statistics • standardized spacings

Access to Document

10.1111/j.1467-9892.2006.00483.x

http://dx.doi.org/10.1111/j.1467-9892.2006.00483.x

Cite this

@article{3401b1150f094c269e6ef2401ec3535a,

title = "Additive Outlier Detection via Extreme-Value Theory.",

abstract = "This article is concerned with detecting additive outliers using extreme value methods. The test recently proposed for use with possibly non-stationary time series by Perron and Rodriguez [Journal of Time Series Analysis (2003) vol. 24, pp. 193–220], is, as they point out, extremely sensitive to departures from their assumption of Gaussianity, even asymptotically. As an alternative, we investigate the robustness to distributional form of a test based on weighted spacings of the sample order statistics. Difficulties arising from uncertainty about the number of potential outliers are discussed, and a simple algorithm requiring minimal distributional assumptions is proposed and its performance evaluated. The new algorithm has dramatically lower level-inflation in face of departures from Gaussianity than the Perron–Rodriguez test, yet retains good power in the presence of outliers.",

keywords = "Additive outliers • extreme order statistics • standardized spacings",

author = "Peter Burridge and Taylor, {A. M. Robert}",

year = "2006",

month = apr,

day = "21",

doi = "10.1111/j.1467-9892.2006.00483.x",

language = "English",

volume = "27",

pages = "685--701",

journal = "Journal of Time Series Analysis",

issn = "0143-9782",

publisher = "Wiley-Blackwell",

number = "5",

}

TY - JOUR

T1 - Additive Outlier Detection via Extreme-Value Theory.

AU - Burridge, Peter

AU - Taylor, A. M. Robert

PY - 2006/4/21

Y1 - 2006/4/21

N2 - This article is concerned with detecting additive outliers using extreme value methods. The test recently proposed for use with possibly non-stationary time series by Perron and Rodriguez [Journal of Time Series Analysis (2003) vol. 24, pp. 193–220], is, as they point out, extremely sensitive to departures from their assumption of Gaussianity, even asymptotically. As an alternative, we investigate the robustness to distributional form of a test based on weighted spacings of the sample order statistics. Difficulties arising from uncertainty about the number of potential outliers are discussed, and a simple algorithm requiring minimal distributional assumptions is proposed and its performance evaluated. The new algorithm has dramatically lower level-inflation in face of departures from Gaussianity than the Perron–Rodriguez test, yet retains good power in the presence of outliers.

AB - This article is concerned with detecting additive outliers using extreme value methods. The test recently proposed for use with possibly non-stationary time series by Perron and Rodriguez [Journal of Time Series Analysis (2003) vol. 24, pp. 193–220], is, as they point out, extremely sensitive to departures from their assumption of Gaussianity, even asymptotically. As an alternative, we investigate the robustness to distributional form of a test based on weighted spacings of the sample order statistics. Difficulties arising from uncertainty about the number of potential outliers are discussed, and a simple algorithm requiring minimal distributional assumptions is proposed and its performance evaluated. The new algorithm has dramatically lower level-inflation in face of departures from Gaussianity than the Perron–Rodriguez test, yet retains good power in the presence of outliers.

KW - Additive outliers • extreme order statistics • standardized spacings

U2 - 10.1111/j.1467-9892.2006.00483.x

DO - 10.1111/j.1467-9892.2006.00483.x

M3 - Article

SN - 0143-9782

VL - 27

SP - 685

EP - 701

JO - Journal of Time Series Analysis

JF - Journal of Time Series Analysis

IS - 5

ER -