Convergence in distribution of almost all sums of independent random elements to a stable law (Q1893636)

Let \(0 < p \leq 2\); \(B\) a separable Banach space with dual space \(B'\); \(Y\), \(Y_ n\), \(n \in N\), independently identically distributed nondegenerate random elements in \(B\); \(\xi\), \(\xi_ n\), \(n \in \mathbb{N}\), independent nondegenerate identically distributed symmetric random variables belonging to the domain of normal attraction of a \(p\)-stable law, moreover the random variables \(Y\) and \(Y_ n\) and the random variables \(\xi\) and \(\xi_ n\) are defined on distinct probability spaces \((\Omega, {\mathfrak A}, P)\) and \((\Omega_ 1, {\mathfrak A}_ 1, P_ 1)\), respectively, while the sequence \(S_ n^ p(\omega) = n^{-1/p} \sum^ n_{i = 1} \xi_ i Y_ i(\omega)\). A criterion for weak convergence of the sequence of distributions \({\mathcal L} (S^ 2_ n(\omega))\) for almost all \(\omega \in \Omega\) to a Gaussian distribution was obtained by \textit{M. Ledoux} and \textit{M. Talagrand} [Probab. Theory Relat. Fields 77, No. 1, 29-47 (1988; Zbl 0617.60008)]. Here we will obtain analogs of these results for the case \(p < 2\).