The existence of moments of the suprema of multiple sums and the strong law of large numbers. (English. Ukrainian original) Zbl 1004.60030

Klesov, O. I.

Theory Probab. Math. Stat. 62, 27-37 (2001); translation from Teor. Jmovirn. Mat. Stat. 62, 27-36 (2000).

Let \(d\geq 1,\) and let \(N^{d}\) be a set of \(d\)-dimensional vectors with natural coordinates. Let us consider the field of i.i.d. random variables \(\{X(\overline n)\); \(\overline n\in N^d\}\) and the field of their sums \(S(\overline n)=\sum_{\overline k\leq\overline n}X(\overline k),\) where “\(\leq\)” means the coordinate order in \(N^{d}.\) The strong law of large numbers in this case is the following relation: \(\lim_{\max}\frac{S(\overline n)}{b(\overline n)}=0\) a.s., where \(\lim_{\max}\) is a limit as \(\max(n_{1},\dots ,n_{d})\to\infty,\) \(\overline n=(n_{1},\dots ,n_{d}),\) and \(\{b(\overline n)\), \(\overline n\in N^{d}\}\) is a field of non-random numbers. Let us also consider the following relation: \[ \lim_{n_{1}\to\infty}\sup_{n_{2}\geq 1,\dots ,n_{d}\geq 1}\frac{|S(\overline n)|}{b(\overline n)}=0 \;\text{ a.s.} \] It is obvious that the first relation follows from the second one. But the inverse assertion is not obvious. The aim of the present paper is to prove the equivalence of these two relations for a special class of fields \(\{b(\overline n),\overline n\in N^{d}\}.\) Another problem discussed in the paper is to find the moment conditions for the existence of the expectation \(E[\sup_{\overline n\in N^{d}}(|S(\overline n)|/b(\overline n))]^q\).

Reviewer: A.V.Swishchuk (Kyïv)

MSC:

60F99	Limit theorems in probability theory
60G50	Sums of independent random variables; random walks
60F15	Strong limit theorems

Keywords:

multiple sums; independent identically distributed random variables; supremum moments; strong law of large number