Inequalities for multivariate infinitely divisible processes

From MaRDI portal





Let \(\{Q(t):t\geq 0\}\) be a one-parameter semigroup of infinitely divisible distributions on the real line. Let \(Z_ A\), \(A\subseteq \{1,...,n\}\), be a family of independent random variables with distribution \(Q_{t(A)}\), where \(t(A)\geq 0\), and define the i th component of a random vector X by \(X_ i=\sum_{i\in A}Z_ A\). Then X is said to have a multivariate infinitely divisible (m.i.d.) distribution with parameter \(t=\{t(A):\emptyset \neq A\subseteq \{1,...,n\}\}\). Now let X,Y have m.i.d. distributions with parameters \(t,t^*.\) The main results of the paper give conditions on \(t,t^*\) for stochastic orderings of the type \[ P(X\geq c)\leq P(Y\geq c)\text{ for all } c\in {\mathbb{R}}^ n. \] Compound Poisson families play a key role. Some applications are given, including inequalities fore Linnik power zones in the scheme of summation of non-identically distributed r.v. Large deviations in terms of Lyapunov fractions are also investigated. It appeared that the probabilities of large deviations take into account mainly not individual properties of summands but the average ones. This was shown first by \textit{W. Wolf} [Math. Nachr. 70, 197-215 (1975; Zbl 0324.60028)] under more strict conditions.











This page was built for publication: Inequalities for multivariate infinitely divisible processes

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1103943)