Characterizing a joint probability distribution by conditionals

SUMMARY We derive conditions under which a set of conditional and marginal probability distributions will uniquely specify an all-positive joint distribution. Our theoretical result may yield insights into the construction and simulation of multivariate probability models.