Discrete semistable probability distributions (Q1122209): Difference between revisions

[For the entire collection see Zbl 0626.00025.] Let \({\mathcal F}=\{f(\cdot,t)\), \(f(0,t)<1\), \(t\geq 0\}\) be a semigroup of generating functions (g.f.) of random variables with values in \(N_ 0=\{0,1,2,...\}\), depending on a continuous parameter, relative to the composition \[ f(\cdot,t_ 1)\circ f(\cdot,t_ 2)\equiv f(f(\cdot,t_ 2),t_ 1)=f(\cdot,t_ 1+t_ 2),\quad t_ 1,t_ 2\geq 0, \] continuous at zero: \[ \lim_{t\downarrow 0}f(z,t)=z,\quad | z| \leq 1. \] A g.f. g of a random variable with values in \(N_ 0\) is said to be \({\mathcal F}\)-semistable if there exist a number \(r>1\) and a g.f. \(f\in {\mathcal F}\) such that \(g=g^ r\circ f.\) The author gives the description of \({\mathcal F}\)-semistable g.f. and some subclasses of these functions; in particular, the canonical representation of \({\mathcal F}\)-semistable g.f., and \({\mathcal F}\)-semistable distributions as limit distributions for distributions of normed sums of independente, identically distributed random variables. \({\mathcal F}\)-semistable distributions and their properties are the analogies of \({\mathcal F}\)-stable distributions and have been investigated by \textit{K. van Harn}, \textit{F. W. Steutel} and \textit{W. Vervaart} in Z. Wahrscheinlichkeitstheor. Verw. Geb. 61, No.1, 97-118 (1982; Zbl 0476.60016). Note that in formula (8) of the text it must be \(q<1\) if \(m>1\) but not \(q>1\).