A strong law for a set-indexed partial sum process with applications to exchangeable and stationary sequences (Q1096959): Difference between revisions

Let \(\{Y_ n\), \(n\geq 1\}\) be a sequence of identically distributed random variables, not necessarily independent nor with finite mean, and let \(\{J_ n\), \(n\geq 1\}\) be an increasing sequence of finite sets of positive integers. Set \(S_ n=\sum_{j\in J_ n}Y_ j\), and let \(\{\) b(n), \(n\geq 1\}\) be any sequence of constants such that \[ n^{- 1}b(n)\to \infty \quad and\quad \sum_{n\geq 1}P[| Y_ 1| >b(n)]<\infty. \] Then, provided an ergodic theorem holds for the partial sums of \(g(| Y_ j|)\), for any \(g\in L_ 1\), \(S_ n/b(| J_ n|)\to 0\) a.s. The result is applied to obtain laws of large numbers for symmetric functions of exchangeable sequences, for stationary sequences and for sequences of pairwise independent random variables.