Group for Research in Decision Analysis

Finite normal mixture copulas for multivariate discrete data modeling

Aristidis K. Nikoloulopoulos Université Laval, Canada

A family of copulas will be introduced which provides flexible dependence structures while being tractable and simple to use for multivariate discrete data modeling. The construction exploits finite mixtures of simple uncorrelated normal distributions. Accordingly, the cumulative distribution is simply the product of univariate normal cumulative functions. At the same time, however, the mixing operation introduces association. The properties of the new family of copulas will be examined and a concrete application will be used to show its applicability.