Abstract
We introduce a family of bivariate discrete distributions whose members are generated by a decreasing mass function p, and with margins given by p. Several properties and examples are obtained, including a family of seemingly novel bivariate Poisson distributions.
Notes
The common pmf is given by for
, with
being the Gauss hypergeometric function. For a ⩾ 1, we can show that p decreases on
and that such a p generates the joint probability mass function in (Equation2
(2) ) via Lemma 1.
A multivariate generalization, which is not pursued here, for a joint distribution for (X1, …Xn) generated from a nonincreasing p, with univariate marginals given by p, as above is of the form
or by a direct evaluation
This is not a necessary condition. For instance, the negative binomial distributions presented earlier in the table are Poisson mixtures with Gamma distributed α.
By taking a sequence of distributions with mean r converging in distribution to a Bernoulli(r).
Note that S is not necessarily unique but it does not matter in such cases which choice is made.