Estimating multivariate extremal dependence: a new proposal

Abstract

Multivariate extreme values require the use of extreme-value copulas, as they appear in the limit of componentwise maxima. These can be characterized by the so-called Pickands dependence function. A new multivariate nonparametric estimator will be presented, along with convergence properties. Based on simulations, we will analyze its performance and compare with well-known estimators from the literature.