Multivariate Rodriguez Copula with Applications

In the appendix, a generalization of Polisicchio distribution (GP) allows to build bivariate distribution by mixing two independent GP’s with a bivariate beta weight distribution. Unfortunately, treating this bivariate distribution is very expensive from a computational point of view. So, bivariate Burr III Rodriguez copula is generalized to multivariate case, studied and applied to some datasets. This copula seems to be very general, analytically manageable and exhibits a parsimonious number of parameters, in comparison to the copulas in literature. Moreover, its restriction to second order interactions seems to be a competitor of the multivariate normal copula. On the other hand, an application shows that second order interaction may be not sufficient to reproduce real data situations.