Publication detail

Sensitivity Assessment and Comparison of Maxima Methods in the Estimation of Extremal Index

HOLEŠOVSKÝ, J.

Original Title

Sensitivity Assessment and Comparison of Maxima Methods in the Estimation of Extremal Index

English Title

Sensitivity Assessment and Comparison of Maxima Methods in the Estimation of Extremal Index

Type

conference paper

Language

en

Original Abstract

Extremal index is the primary measure of local dependence of extreme values and plays important role in extreme value estimation for stationary processes. The maxima estimators are often preferred in practical situations. These estimators, based on properties of the block maxima, are asymptotically characterized by the Generalized extreme value distribution. In contrast to other methods, the maxima estimators gain advantage in stability to the choice of auxiliary parameters. Still the main part of the maxima methods is selection of a proper approximation to the marginal distribution of the underlying process. Although the suitability of the approximation may significantly affect the estimation quality, to the effect of available approaches has not been paid a great interest in the literature. The aim of this contribution is the comparison of available sampling schemes and the assessment of sensitivity of existing maxima estimates of the extremal index.

English abstract

Extremal index is the primary measure of local dependence of extreme values and plays important role in extreme value estimation for stationary processes. The maxima estimators are often preferred in practical situations. These estimators, based on properties of the block maxima, are asymptotically characterized by the Generalized extreme value distribution. In contrast to other methods, the maxima estimators gain advantage in stability to the choice of auxiliary parameters. Still the main part of the maxima methods is selection of a proper approximation to the marginal distribution of the underlying process. Although the suitability of the approximation may significantly affect the estimation quality, to the effect of available approaches has not been paid a great interest in the literature. The aim of this contribution is the comparison of available sampling schemes and the assessment of sensitivity of existing maxima estimates of the extremal index.

Keywords

extreme value; extremal index; stationary series; block maxima; resampling

Released

18.12.2017

Publisher

University of Defence

Location

Brno, Czech Republic

ISBN

978-80-7582-026-6

Book

MITAV 2017, Post-Conference Proceedings of Extended Versions of Selected Papers

Pages from

110

Pages to

120

Pages count

10

URL

BibTex


@inproceedings{BUT142575,
  author="Jan {Holešovský}",
  title="Sensitivity Assessment and Comparison of Maxima Methods in the Estimation of Extremal Index",
  annote="Extremal index is the primary measure of local dependence of extreme values and plays important role in extreme value estimation for stationary processes. The maxima estimators are often preferred in practical situations. These estimators, based on properties of the block maxima, are asymptotically characterized by the Generalized extreme value distribution. In contrast to other methods, the maxima estimators gain advantage in stability to the choice of auxiliary parameters. Still the main part of the maxima methods is selection of a proper approximation to the marginal distribution of the underlying process. Although the suitability of the approximation may significantly affect the estimation quality, to the effect of available approaches has not been paid a great interest in the literature. The aim of this contribution is the comparison of available sampling schemes and the assessment of sensitivity of existing maxima estimates of the extremal index.",
  address="University of Defence",
  booktitle="MITAV 2017, Post-Conference Proceedings of Extended Versions of Selected Papers",
  chapter="142575",
  howpublished="electronic, physical medium",
  institution="University of Defence",
  year="2017",
  month="december",
  pages="110--120",
  publisher="University of Defence",
  type="conference paper"
}