Maximal inequalities and convergence results on multidimensionally indexed demimartingales (Q6151245): Difference between revisions

A collection of multidimensionally indexed random variables \(\{S_\mathbf{n}:\mathbf{n}\in\mathbb{N}^k\}\) is said to be a multi-indexed demimartingale if \[ E\left[(S_\mathbf{j}-S_\mathbf{i})f(S_\mathbf{k},\mathbf{k}\leq\mathbf{i})\right]\geq0 \] for all \(\mathbf{i},\mathbf{j}\in\mathbb{N}^k\) with \(\mathbf{i}\leq\mathbf{j}\) and all componentwise non-decreasing functions \(f\), where we write \((i_1,\ldots,i_k)\leq(j_1,\ldots,j_k)\) if \(i_l\leq j_l\) for all \(l=1,\ldots,k\). This generalises the class of multi-indexed martingales, and partial sums of associated random variables with zero mean. In the present paper the authors establish a selection of inequalities for multi-indexed demimartingales, including Doob-type inequalities and maximal results, a Chow-type inequality, maximal inequalities in terms of Orlicz functions, an upcrossing inequality, and a Whittle-type inequality.