Peaks Over Thresholds Modeling With Multivariate Generalized Pareto Distributions

Anna Kiriliouk, Holger Rootzén, Johan Segers, Jennifer L. Wadsworth

Research output: Contribution to journalArticle

Abstract

When assessing the impact of extreme events, it is often not just a single component, but the combined behaviour of several components which is important. Statistical modelling using multivariate generalized Pareto (GP) distributions constitutes the multivariate analogue of univariate peaks over thresholds modelling, which is widely used in finance and engineering. We develop general methods for construction of multivariate GP distributions and use them to create a variety of new statistical models. A censored likelihood procedure is proposed to make inference on these models, together with a threshold selection procedure, goodness-of-fit diagnostics, and a computationally tractable strategy for model selection. The models are fitted to returns of stock prices of four UK-based banks and to rainfall data in the context of landslide risk estimation. Supplementary materials and codes are available online.
Original languageEnglish
Pages (from-to)123-135
JournalTechnometrics
Volume61
Issue number1
DOIs
Publication statusPublished - 2 Jan 2019
Externally publishedYes

Fingerprint

Peaks over Threshold
Generalized Pareto Distribution
Modeling
Extreme Events
Landslide
Statistical Modeling
Stock Prices
Landslides
Selection Procedures
Rainfall
Finance
Goodness of fit
Model Selection
Statistical Model
Univariate
Rain
Likelihood
Diagnostics
Engineering
Analogue

Keywords

  • Financial risk
  • Landslides
  • Multivariate extremes
  • Tail dependence

Cite this

Kiriliouk, Anna ; Rootzén, Holger ; Segers, Johan ; Wadsworth, Jennifer L. / Peaks Over Thresholds Modeling With Multivariate Generalized Pareto Distributions. In: Technometrics. 2019 ; Vol. 61, No. 1. pp. 123-135.
@article{a6e8538d86c24a12a423b1afcd0c2fbd,
title = "Peaks Over Thresholds Modeling With Multivariate Generalized Pareto Distributions",
abstract = "When assessing the impact of extreme events, it is often not just a single component, but the combined behaviour of several components which is important. Statistical modelling using multivariate generalized Pareto (GP) distributions constitutes the multivariate analogue of univariate peaks over thresholds modelling, which is widely used in finance and engineering. We develop general methods for construction of multivariate GP distributions and use them to create a variety of new statistical models. A censored likelihood procedure is proposed to make inference on these models, together with a threshold selection procedure, goodness-of-fit diagnostics, and a computationally tractable strategy for model selection. The models are fitted to returns of stock prices of four UK-based banks and to rainfall data in the context of landslide risk estimation. Supplementary materials and codes are available online.",
keywords = "Financial risk, Landslides, Multivariate extremes, Tail dependence",
author = "Anna Kiriliouk and Holger Rootz{\'e}n and Johan Segers and Wadsworth, {Jennifer L.}",
year = "2019",
month = "1",
day = "2",
doi = "10.1080/00401706.2018.1462738",
language = "English",
volume = "61",
pages = "123--135",
journal = "Technometrics",
issn = "0040-1706",
publisher = "American Statistical Association",
number = "1",

}

Peaks Over Thresholds Modeling With Multivariate Generalized Pareto Distributions. / Kiriliouk, Anna; Rootzén, Holger; Segers, Johan; Wadsworth, Jennifer L.

In: Technometrics, Vol. 61, No. 1, 02.01.2019, p. 123-135.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Peaks Over Thresholds Modeling With Multivariate Generalized Pareto Distributions

AU - Kiriliouk, Anna

AU - Rootzén, Holger

AU - Segers, Johan

AU - Wadsworth, Jennifer L.

PY - 2019/1/2

Y1 - 2019/1/2

N2 - When assessing the impact of extreme events, it is often not just a single component, but the combined behaviour of several components which is important. Statistical modelling using multivariate generalized Pareto (GP) distributions constitutes the multivariate analogue of univariate peaks over thresholds modelling, which is widely used in finance and engineering. We develop general methods for construction of multivariate GP distributions and use them to create a variety of new statistical models. A censored likelihood procedure is proposed to make inference on these models, together with a threshold selection procedure, goodness-of-fit diagnostics, and a computationally tractable strategy for model selection. The models are fitted to returns of stock prices of four UK-based banks and to rainfall data in the context of landslide risk estimation. Supplementary materials and codes are available online.

AB - When assessing the impact of extreme events, it is often not just a single component, but the combined behaviour of several components which is important. Statistical modelling using multivariate generalized Pareto (GP) distributions constitutes the multivariate analogue of univariate peaks over thresholds modelling, which is widely used in finance and engineering. We develop general methods for construction of multivariate GP distributions and use them to create a variety of new statistical models. A censored likelihood procedure is proposed to make inference on these models, together with a threshold selection procedure, goodness-of-fit diagnostics, and a computationally tractable strategy for model selection. The models are fitted to returns of stock prices of four UK-based banks and to rainfall data in the context of landslide risk estimation. Supplementary materials and codes are available online.

KW - Financial risk

KW - Landslides

KW - Multivariate extremes

KW - Tail dependence

UR - http://www.scopus.com/inward/record.url?scp=85049077358&partnerID=8YFLogxK

U2 - 10.1080/00401706.2018.1462738

DO - 10.1080/00401706.2018.1462738

M3 - Article

AN - SCOPUS:85049077358

VL - 61

SP - 123

EP - 135

JO - Technometrics

JF - Technometrics

SN - 0040-1706

IS - 1

ER -