A generalized Sibuya distribution (Q1656865)

The authors study the positive integer-valued random variable \(N\) defined by \[ \mathbb{P}(N=n)=\left(1-\frac{\alpha}{\nu+1}\right)\cdots\left(1-\frac{\alpha}{\nu+n-1}\right)\frac{\alpha}{\nu+n}\,, \] for \(n=1,2,\ldots\), and with parameters \(\nu\geq0\) and \(0<\alpha<\nu+1\). This arises as the waiting time for the first success in a sequence of independent Bernoulli trials \(I_1,I_2,\ldots\) whose success probabilities are given by \(\mathbb{P}(I_j=1)=\frac{\alpha}{\nu+j}\). This distribution is termed the generalized Sibuya distribution. The authors show that it may also be constructed as a randomly stopped Poisson process, or as a certain mixed geometric distribution. The majority of the paper is devoted to fundamental results on the generalized Sibuya distribution, such as moments, tail behaviour, generating functions, divisibility and decomposibility. Some attention is also given to statistical aspects, including method-of-moments and maximum likelihood estimation of the parameters. The authors also present some results on the Sibuya distribution (the special case \(\nu=0\)). This includes some comments on random maxima and minima, a characterization in terms of a pure death process, and a random process whose marginals are Sibuya-distributed.