Representation of multivariate Bernoulli distributions with a given set of specified moments

We propose a simple but new method of characterizing multivariate Bernoulli variables belonging to a given class, i.e., with some specified moments. Within a given class, this characterization allows us to generate easily a sample of mass functions. It also provides the bounds that all the moments must satisfy to be compatible and the possibility to choose the best distribution according to a certain criterion. For the special case of the Fréchet class of the multivariate Bernoulli distributions with given margins, we find a polynomial characterization of the class. Our characterization allows us to have bounds for the higher order moments. An algorithm is presented and illustrated.