On the Hsu-Robbins-Erdős-Spitzer-Baum-Katz theorem for random fields (Q645407)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | On the Hsu-Robbins-Erdős-Spitzer-Baum-Katz theorem for random fields |
scientific article |
Statements
On the Hsu-Robbins-Erdős-Spitzer-Baum-Katz theorem for random fields (English)
0 references
15 November 2011
0 references
The classical results on convergence-rates in the law of large numbers involving the sums \(\sum\limits_{n=1}^\infty n^{\alpha r-2}\mathrm{P}\{|S_n|>n^{\alpha}_r\}\), where \(r>0\), \(\alpha>\frac{1}{2}\) and \(\alpha r\geq 1\), were extended to the case \(\alpha=\frac{1}{2}\) by adding additional logarithms. These problems have been generalized to random fields by the first author [``Marcinkiewicz laws and convergence rates in the law of large numbers for random variables with multidimensional indices'', Ann. Probab. 6, 469--482 (1978; Zbl 0383.60030); ``Convergence rates for probabilities of moderate deviations for sums of random variables with multidimensional indices'', ibid. 8, 298--313 (1980; Zbl 0429.60022)] in case that \(\alpha\) is the same in all directions of the field. The present paper deals with the case when \(\alpha\) differs in different directions. Let \({\mathbb{Z}}^d_+\), \(d\geq2\), denote the positive \(d\)-dimensional integer lattice with coordinate-wise partial ordering \(\leq\). For \(n=(n_1, \ldots, n_d)\), denote \(|n|=\prod\limits_{n=1}^k n_k\), and \(n\rightarrow \infty\) means \(n_k \rightarrow \infty\) for all \(k=1,2,\ldots,d\). In this paper, different powers are allowed for different coordinates. For \(\pmb{\alpha}=(\alpha_1,\alpha_2,\ldots, \alpha_d)\), the coordinates are arranged in non-decreasing order, namely, \(\frac{1}{2}\leq \alpha_1\leq\alpha_2\leq\ldots\leq\alpha_d<1\). Furthermore, \(\mathbf{n^{\alpha}}=(n^{\alpha_1}_1,n^{\alpha_2}_2,\ldots,n^{\alpha_d}_d)\); \[ |\mathbf{n^{\alpha}}|=\prod\limits_{k=1}^{d}n^{\alpha_k}_k, \quad p=\max\left\{k:\alpha_k=\alpha_1\right\}. \] Section 1 contains the results of the paper without proofs. The main statement is Theorem 1.3. Let \(r\geq0\), \(\alpha_1>\frac{1}{2}\) and \(\alpha_1 r\geq1\), let \(\{X_k,k\in{\mathbb{Z}}^d_+\}\) be i.i.d. random variables with partial sums \(S_\mathbf{n}=\sum\limits_{\mathbf{k}\leq \mathbf{n}}X_\mathbf{k}\), \(\mathbf{n}\in{\mathbb{Z}}^d_+\). If \(\sum |X|^r(\log^+|X|)^{p-1}<\infty\) and if \(r\geq 1\), \(\operatorname{E}X=0\), then, for all \(\varepsilon>0\), \[ \sum\limits_{\mathbf{n}}|\mathbf{n}|^{\alpha_1 r-2}\mathrm{P}\left\{|S_\mathbf{n}|>|\mathbf{n^{\alpha}}|\varepsilon\right\}<\infty; \] \[ \sum\limits_\mathbf{n}|\mathbf{n}|^{\alpha_1 r-2}\mathrm{P}\left\{\max\limits_{\mathbf{k}\leq \mathbf{n}}|S_\mathbf{k}|>|\mathbf{n^{\alpha}}|\varepsilon\right\}<\infty. \] If \(\alpha_1 n>1\), then \[ \sum\limits_{j=1}^{\infty}j^{\alpha_1 r-2}\mathrm{P}\left\{\sup\limits_{j\leq |\mathbf{k}|}|S_\mathbf{k}/|\mathbf{k^{\alpha}}||>\varepsilon\right\}<\infty \] for all \(\varepsilon>0\). Conversely, if one of the sums is finite for all \(\varepsilon>0\), then \[ \operatorname{E}|X|^r(\log^+|X|)^{p-1}<\infty \] if \(r\geq1\), \(\operatorname{E}X=0\). The next statements (Theorems 1.4, 1.5, 1.6 and 1.7) involve conditions on \(|X|\) with added logarithms and/or iterated ones. In Section 2, auxiliary tools are proven, expect for Lemma 2.2 which is proven in the Appendix. Then, in Section 3, Theorem 1.3 is proven, while the other theorems are proven in Sections 4 and 5. Section 6 contains some observations on the statements and methods of proving used. Some analogues of the ``last exit times'' are discussed and two theorems on their moments are presented. Two estimates of the counting variable for laws of large numbers and for multiindex laws of iterated logarithms are mentioned as well.
0 references
sums of i.i.d. random variables
0 references
random fields
0 references
law of large numbers
0 references
law of the iterated logarithm
0 references
convergence rates
0 references
last exit time
0 references
0 references