Improved B-spline inference on bivariate extreme-value copulas

Abstract

In this talk we describe a B-spline estimator of a generator (Pickands dependence function) of a bivariate extreme-value copula. The estimator is built upon an A-plot -- a recently introduced graphical tool for assessing whether the underlying dependence structure displays extreme tail behavior. An extension to higher order B-splines requires more subtle implementation yet allows for a transition to spectral measure estimation of the underlying extreme-value copula or the extreme-value attractor. The proposed estimator performs well and compares good to the existing methods. This is a joint work with Christian Genest and Johana Neslehova.