Peaks Over Thresholds Modeling With Multivariate Generalized Pareto Distributions

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

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

47 Citations (Scopus)

Abstract

When assessing the impact of extreme events, it is often not just a single component, but the combined behavior of several components which is important. Statistical modeling using multivariate generalized Pareto (GP) distributions constitutes the multivariate analogue of univariate peaks over thresholds modeling, 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
Number of pages13
JournalTechnometrics
Volume61
Issue number1
DOIs
Publication statusPublished - 25 Jun 2018

Bibliographical note

Funding Information:
The authors gratefully acknowledge support from: the Knut and Alice Wallenberg foundation (Kiriliouk, Rootzén, Wadsworth); “Projet d’Actions de Recherche Concertées” No. 12/17-045 of the “Communauté française de Belgique” (Kiriliouk, Segers); IAP research network grant P7/06 of the Belgian government (Segers); EPSRC fellowship grant EP/P002838/1 (Wadsworth). Finally, we thank the Abisko Scientific Research Station for access to their rainfall data.

Publisher Copyright:
© 2018, © 2018 The Author(s). Published with license by Taylor & Francis.

Fingerprint

Dive into the research topics of 'Peaks Over Thresholds Modeling With Multivariate Generalized Pareto Distributions'. Together they form a unique fingerprint.

Cite this