Sharp inequalities involving multiplicative chaos sums (Q6607779)

According to the author's abstract: ``The present note is an addition to the author's recent paper \textit{G. A. Karagulyan} [Proc. Am. Math. Soc. 149, No. 9, 3725--3737 (2021; Zbl 1469.42027)], concerning general multiplicative systems of random variables. Using some terms and methodology of that paper, we obtain a general extremal inequality, with corollaries involving Rademacher chaos sums. In particular we prove that a system of functions generated by bounded products of a multiplicative system is a convergent system.'' \par The main result of the paper is given in:\par Theorem 1.1.: Let a system of random variables \(\phi=\{\phi_k\,:\, k=1,\ldots, n\}\) satisfy \(A_k\le\phi_k\le B_k\), where \(A_k<0<B_k\), \(d\leq n\) and let \(G\,:\,\mathbb{R}^n\rightarrow \mathbb{R}^{+}\) be a separately convex function. Then the inequality \[\mathrm{E}[G(\phi_1,\ldots, \phi_n)]\leq (1+\mu_d(\phi))\mathbb{E}[G(\xi_1,\ldots, \xi_n)]\] holds, where \(\xi_k\,k=1,2,\ldots, n\), are \(\{A_k,B_k\}\)-valued \(d\)-independent mean zero variables. If the system \(\phi\) is \(d\)-multiplicative (that is \(\mu_d(\phi)=0\)), then the inequality holds with arbitrary (not necessarily positive) separately convex function \(G\).\par Here \(\mathbb{Z}_n=\{1,\ldots, n\}\) and \(\mathbb{Z}_{\infty}=\{1,2,\ldots\}\). Given \(d\leq n\leq \infty\), the \(d\)-multiplicative error of the system \(\phi\) is the quantity \[\mu_d(\phi)=\sum_{A\subset\mathbb{Z}_n,\# (A)\leq d,j}\,\prod_{j\in A}\, (C_j)^{-1}\left\vert \mathrm{E} [\prod_{j\in A}\,\phi_j] \right\vert ,\] where \(C_k=\min \{A_k,B_k\}\).