A strong law of large numbers for partial-sum processes indexed by sets (Q796170)

Let \(J={\mathbb{N}}^ d\) and let \(\{X_ j:j\in J\}\) be a family of independent, identically distributed random variables with \(E| X_ j|<\infty\) and \(EX_ j=\mu\). For any Borel measurable set \(B\subseteq [0,\infty)^ d\) let \(| B| =Lebesgue\) measure of B and let \(S(B)=\sum_{j\in B}X_ j\), the partial sum process indexed by sets. Let \(B(\delta)=\{x: dist(x,\partial B)<\delta \}\) be the \(\delta\) annulus about \(\partial B\), the boundary of B. The following law of large numbers is proved: Let \({\mathcal A}\) be a collection of Borel measurable subsets of \([0,1]^ d\) such that \(r(\delta)\equiv \sup_{A\in {\mathcal A}}| A(\delta)| \to 0\) as \(\delta \to 0\). Then \(\sup_{A\in {\mathcal A}}| S(nA)/n^ d-\mu | A| | \to 0\) as \(n\to \infty\). In particular the theorem holds if \({\mathcal A}\) is the collection of all convex sets of \([0,\infty)^ d\). This is a nice complement to the results of Bass and Pyke, Alexander and Pyke, Morrow and Philipp, and others, on the CLT for partial sum processes indexed by sets.