On Rio's proof of limit theorems for dependent random fields (Q6204185)

Letting \(\{X_{m,n}:m\geq1,n\geq1\}\) be a doubly-indexed array of random variables, one of the main results of the present paper is a Hsu-Robbins-Erdős-Spitzer-Baum-Katz theorem for the maximum of double sums of the \(X_{m,n}\) under the assumption that there exists a real number \(q\geq1\) such that for all finite subsets \(I\subseteq\mathbb{N}^2\) and all families of increasing functions \(\{f_\lambda:\lambda\in I\}\) there exists a finite constant \(C(q)\) depending only on \(q\) such that \[ \mathbb{E}\left|\sum_{\lambda\in I}\left[f_\lambda(X_\lambda)-\mathbb{E}f_\lambda(X_\lambda)\right]\right|^{2q}\leq C(q)\left(|I|\max_{\lambda\in I}\mathbb{E}|f_\lambda(X_\lambda)|^{2q}+|I|^q\max_{\lambda\in I}\left[\mathbb{E}f_\lambda^2(X_\lambda)\right]^q\right)\,, \] when these expectations are finite. The author also establishes a Feller weak law of large numbers and a Pyke-Root theorem for mean convergence of the maximum of double sums of the \(X_{m,n}\) under the above condition with \(q=1\). The proofs use extensions to random fields of methods developed by \textit{E. Rio} [C. R. Acad. Sci., Paris, Sér. I 320, No. 4, 469--474 (1995; Zbl 0832.60042)], with the author giving a new Rosenthal-type inequality for double sums of dependent random variables using Rio's technique as a warm-up for the main results of the paper.