Probability distributions with binomial moments

Abstract

We prove that if $p\geq 1$ and $-1\leq r\leq p-1$ then the binomial sequence $\binom{np+r}{n}$, $n=0,1,...$, is positive definite and is the moment sequence of a probability measure $\nu(p,r)$, whose support is contained in $\left[0,p^p(p-1)^{1-p}\right]$. If $p>1$ is a rational number and $-1<r\leq p-1$ then $\nu(p,r)$ is absolutely continuous and its density function $V_{p,r}$ can be expressed in terms of the Meijer $G$-function. In particular cases $V_{p,r}$ is an elementary function. We show that for $p>1$ the measures $\nu(p,-1)$ and $\nu(p,0)$ are certain free convolution powers of the Bernoulli distribution. Finally we prove that the binomial sequence $\binom{np+r}{n}$ is positive definite if and only if either $p\geq 1$, $-1\leq r\leq p-1$ or $p\leq 0$, $p-1\leq r \leq 0$. The measures corresponding to the latter case are reflections of the former ones.