Synthèse de processus bivariés non gaussiens à mémoire longue

A procedure is proposed for the synthesis of bivariate stochastic processes whose marginal distributions and autoand intercovariance functions are a priori prescribed. Calculations corresponding to the case of marginal gamma distributions (with scale parameter restricted to multiple of semi integer values) are detailed. This procedure is illustrated at work for the synthesis of long range dependent processes whose bivariate components possess the so-called fractal connectivity property. The potential extension to other distributions is commented, together with the difficulties encountered when trying to obtain processes whose components have marginal distributions chosen amongst different families of laws.