Title: On the estimation of elliptical copula generators
Authors: Alexis Derumigny - University of Twente (Netherlands) [presenting]
Jean-David Fermanian - Ensae-Crest (France)
Abstract: The problem of estimating elliptical copula models is studied. They are defined by a correlation matrix and an unknown function (the ``generator''). Although the estimation of the correlation matrix has been well-studied, estimating the generator is much harder. This model is in fact not identifiable, in the sense that different generators can correspond to the same copula. Therefore, we propose a possible identification constraint. We present an iterative algorithm to nonparametrically estimate the generator, using kernel-smoothing. Finally, we give some numerical results on simulated and real data.