scientific article; zbMATH DE number 2133883

From MaRDI portal
Publication:3159103





Operator \(\nu\)-stable laws \(\lambda\) on a finite-dimensional vector space are defined as limit distributions of randomized sums, \(S_{N_n} := A_n\sum_1^{N_n} \left(X_i - b_n\right)\to Y\), where \(Y\) is \(\lambda\)-distributed, \((X_i, N_n)\) are independent, furthermore, the r.v.s \(X_i\) are assumed to be iid and the random times \(N_n\) fulfil \(N_n \to \infty\) and \(N_n / n \to Z\), a positive r.v. with distribution \(\nu\). If the r.v.s \(X_i\) belong to the domain of attraction of some operator stable law \(\omega\), then well-known transfer theorems, the first due to B. V. Gnedenko, yield the representation \[ \lambda = \int_0^{\infty} \omega^t \,d\nu(t).\tag \(*\) \] Under a technical condition, namely that \(\omega\) in \(*\) is identified by the mixing measure \(\nu\), the authors prove in Theorem 3.3 that the domain of operator attraction of \(\omega\) and the (suitably defined) domain of \(\nu\)-attraction of \(Y\) resp. of \(\lambda\) coincide. The most prominent examples are (operator-)geometric-stable laws -- a generalization of (classical) geometric stable laws -- which are shortly described in Section 6. For more details see, e.g, the authors and \textit{A. Pandorska} [``Operator geometric stable laws, J. Multivariate Anal. (to appear)]. Remark. Transfer theorems, geometric-divisible and in particular -stable laws were investigated in the past also for more general convolution structures. For investigations on groups and their tangent spaces see e.g. \S \ 2.13 in the monograph by \textit{W. Hazod} and \textit{E. Siebert} [``Stable probability measures on Euclidean spaces and on locally compact groups (2001; Zbl 1002.60002)], which includes also a survey on the literature.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3159103)