Title: Modeling time-changing joint extremes via Bernstein polynomials
Authors: Timothy Hanson - Medtronic (United States) [presenting]
Miguel de Carvalho - School of Mathematics, University of Edinburgh (Portugal)
Abstract: A Bayesian model for monitoring the dynamics ruling the dependence between extreme values over time is developed and validated. The model is based on a mean-constrained Bernstein polynomial that can be used to induce a prior on the space of all time-changing extreme value copulas. Practical notes on obtaining and interpreting inference are presented. We examine the performance of the model via a simulation study, and illustrate the methods with a real data case study.