Multivariate Nonstationary Extremes

Miguel de Carvalho (University of Edinburgh, UK)

In this talk, I will showcase recent developments on statistical methods for non-stationary extremal dependence structures. Working within a framework of asymptotic dependence, I introduce regression models for the angular density of multivariate extreme value distributions that can be used to track how extremal dependence evolves over time. The proposed methods are tailored for modelling the dynamics of the synchronization of the extreme values of a random vector over time, or another covariate of interest. The methods are illustrated with data from Finance and Climatology.