Reparameterization of extreme value framework for improved Bayesian workflow - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Journal Articles Computational Statistics and Data Analysis Year : 2023

Reparameterization of extreme value framework for improved Bayesian workflow

Abstract

Using Bayesian methods for extreme value analysis offers an alternative to frequentist ones, with several advantages such as easily dealing with parametric uncertainty or studying irregular models. However, computation can be challenging and the efficiency of algorithms can be altered by poor modelling choices, and among them the parameterization is crucial. We focus on the Poisson process characterization of univariate extremes and outline two key benefits of an orthogonal parameterization. First, Markov chain Monte Carlo convergence is improved when applied on orthogonal parameters. This analysis relies on convergence diagnostics computed on several simulations. Second, orthogonalization also helps deriving Jeffreys and penalized complexity priors, and establishing posterior propriety thereof. Our framework is applied to return level estimation of Garonne flow data (France).
Fichier principal
Vignette du fichier
draft_Poisson_parameterisation-5 (1).pdf (3.89 Mo) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-03806159 , version 1 (07-10-2022)
hal-03806159 , version 2 (27-01-2023)
hal-03806159 , version 3 (26-04-2023)
hal-03806159 , version 4 (09-06-2023)

Licence

Attribution

Identifiers

Cite

Théo Moins, Julyan Arbel, Stéphane Girard, Anne Dutfoy. Reparameterization of extreme value framework for improved Bayesian workflow. Computational Statistics and Data Analysis, 2023, 187, pp.1-21. ⟨10.1016/j.csda.2023.107807⟩. ⟨hal-03806159v4⟩
161 View
129 Download

Altmetric

Share

Gmail Facebook X LinkedIn More