Séminaire de statistique, Université Laval, 13 octobre 2011

le 10 octobre 2011 à 11:14
Jean-Francois Plante

Le jeudi 13 octobre 2011, à 13 h 30, en la salle 2512 du pavillon Adrien-Pouliot, Université Laval.

Inference for general pair-copula constructions
Elif Acar, Université McGill

Pair-copula constructions (PCCs) allow construction of highly flexible multivariate copula models by using only bivariate copulas as building blocks. A standard assumption, made mainly for convenience of inference, is that the conditional pair-copulas of a PCC depend on the conditioning variable(s) only through their conditional margins. In this talk, I will give a brief review of inference techniques for simplified PCCs and present an alternative nonparametric smoothing methodology that relaxes the simplifying assumption. Model selection for PCCs will also be addressed within the proposed method.