Open AccessA Different Way of Choosing a Threshold in a Bivariate Extreme Value Study
Author(s) -
Andréhette Verster,
Nicholas Kwaramba
Publication year - 2021
Publication title -
statistics, optimization and information computing
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.297
H-Index - 12
eISSN - 2311-004X
pISSN - 2310-5070
DOI - 10.19139/soic-2310-5070-1318
Subject(s) - bivariate analysis , univariate , extreme value theory , threshold limit value , threshold model , variance (accounting) , viewpoints , bayesian probability , econometrics , value (mathematics) , statistics , computer science , multivariate statistics , mathematics , economics , physics , medicine , environmental health , accounting , acoustics
The choice of optimum threshold in Extreme Value Theory, peaks over threshold, has been a topic of discussionfor decades. A threshold must be chosen high enough to control the bias of the extreme value index. On the other hand, if a threshold is chosen too high the variance becomes a problem. This is a very difficult trade-off and has been studied over the years from various viewpoints. More often these studies aim at methods for choosing the threshold in univariate settings. Not as many literature are available for choosing the threshold in a multivariate setting. In this paper we consider an approach for choosing the threshold when working with bivariate extreme values above a threshold. This approach makes use of Bayesian methodology. It adds value to the existing literature since it is also possible to use this approach without visual inspection.