Inequalities for multivariate compound Poisson distributions (Q1110180)

Let \(\{Q_ t\); \(t\geq 0\}\) be a one-parameter Poisson family. Let \({\mathcal A}=\{A:\) \(A\subseteq \{1,2,...,n\}\}\), n positive integer, and t(A) be a number associated with \(A\in {\mathcal A}\). Let \(\{Z_ A:\) \(A\in {\mathcal A}\}\) be independent random variables with \(Z_ A\) distributed according to \(Q_{t(A)}\). Define \(x=(x_ 1,...,x_ n)\) by \(x_ i=\sum_{A:i\in A}Z_ A\), \(i=1,2,...,n\). Then x is said to have a multivariate compound Poisson distribution with parameter \(t=\{t(A):\) \(\emptyset \neq A\in {\mathcal A}\}\) based on the family \(Q_ t.\) Let Y have a multivariate compound Poisson distribution \(t^*=\{t^*(A):\) \(\emptyset \neq A\in {\mathcal A}\}\) based on \(Q_ t\). Under these assumptions, the author has established the following inequalities: \[ (a)\quad P\{x\geq c\}\leq P\{Y\geq c\}\quad for\quad all\quad c\in {\mathbb{R}}^ n\Rightarrow \sum_{A:A\supseteq B}t(A)\leq \sum_{A:A\supseteq B}t^*(B)\quad for\quad all\quad B\neq \emptyset,\quad B\in {\mathcal A}. \] \[ (b)\quad P\{x\leq c\}\leq P\{Y\leq c\}\quad for\quad all\quad c\in {\mathbb{R}}^ n\Rightarrow \sum_{A:A\cap B\neq \emptyset}t(A)\geq \sum_{A:A\cap B\neq \emptyset}t^*(A)\quad for\quad all\quad B\neq \emptyset,\quad B\in {\mathcal A}. \] Part (a) is a converse to a theorem of \textit{L. O. Brown} and \textit{Y. Rinott} [see ibid. 642-657 (1988)] and part (b) has already been proved by Brown and Rinott under general set up but the proof presented here is in the spirit of Part (a).