Exact strong laws for multidimensionally indexed random variables (Q5935556): Difference between revisions

Let \(\{X,X_{{\mathbf n}}\), \({\mathbf n} \in \mathbb{Z}_+^d\}\) be i.i.d. random variables. The problem of finding norming constants \(\{a_{{\mathbf n}}\), \({\mathbf n} \in \mathbb{Z}_+^d\}\) and \(\{b_N, N \geq 1\}\) to guarantee that there exists a.s. \(S = \lim_{N \to \infty} \frac{1}{b_N}\sum_{|{\mathbf n}|\leq N} a_{{\mathbf n}} X_{{\mathbf n}}\) is studied. Here \(|{\mathbf n}|= n\) for \({\mathbf n} \in A_n\) where \(\{A_n, n \geq 1\}\) is a partition of \(\mathbb{Z}_+^d\) into nonempty disjoint sets. The cases \(EX = 0\) and \(E|X|= \infty\) are examined (under the hypothesis that \(x P(|X|> x) \approx L(x)\) as \(x \to \infty\) where \(L(x)\) is a slowly varying function). In both cases a single condition of the form \(\sum_{n=1}^\infty d_nP(|X|> c_n) < \infty\) is used where \(\{d_n\}\) and \(\{c_n\}\) are specified auxiliary sequences. Some previous results [see, e.g., the author, Bull. Inst. Math., Acad. Sin. 23, No. 4, 335-341 (1995; Zbl 0835.60026) and. ibid 28, No. 3, 141-166 (2000; Zbl 0966.60024)] are extended. An explicit formula for \(S\) is given involving the ``parameter of symmetry'' of the \(X\) distribution. Three examples are considered as well (comprising, e.g., the St. Petersburg game).