The domain of partial attraction of a Poisson law (Q2640985): Difference between revisions

Let \(X_ 1,X_ 2,..\). be a sequence of independent and identically distributed real-valued random variables with common distribution function F. \textit{A. Groshev} [Bull. Acad. Sci. URSS, Sér. Math. 5, 165-172 (1941; Zbl 0024.42504)] gave a characterization for F belonging to the domain of partial attraction of some Poisson law, i.e., for the existence of some \(\lambda >0\), a subsequence \(n'\) of the sequence of all positive integers and two sequences \(A_{n'}>0\) and \(C_{n'}\) of real constants such that \[ A^{-1}_{n'}\{\sum^{n'}_{i=1}X_ i- C_{n'}\}\to^{{\mathcal D}}Y_{\lambda}\quad as\quad n'\to \infty, \] where \(\to^{{\mathcal D}}\) denotes convergence in distribution and \(Y_{\lambda}\) is a Poisson random variable with mean \(\lambda\). The present paper provides a characterization of the domain of partial attraction of a Poisson law with given mean \(\lambda >0\). The results also identify the portion of the sample \(X_ 1,...,X_{n'}\) that contributes the limiting Poisson behavior of the sum and disclose the effect of the extreme summands. It turns out that the limiting Poisson distributions arise in two qualitatively different ways: For \(\lambda >1\) the norming factors \(A_{n'}\) are determined by a truncated variance function of F, whereas for \(\lambda <1\) they are necessarily of a different order of magnitude. In the borderline case \(\lambda =1\) both possibilities may occur, as shown by an example.