On the asymptotic behavior of increments of sums of independent random variables. (Q1432293)

Let \(S_n\) be the \(n\)th partial sum of i.i.d. nondegenerate real-valued random variables with positive finite expectation. The author considers the almost sure limit behaviour of maximal increments \(\max_{0\leq k\leq n-a_n}(S_{k+a_n}-S_k)\) and related statistics, where \(1\leq a_n\leq n\) is an increasing sequence of integers. Sufficient conditions for almost sure upper and lower bounds of the normalized statistics, as well as general results for the almost sure limits are given. Since no proofs are presented, the conditions remain unclear, though very likely coming from large deviation techniques. Depending on the asymptotic behaviour of \(a_n/\log n\), various classical limit theorems arise from the general results. Among these, the author derives Erdős-Rényi type laws, strong laws of large numbers, and one-sided laws of the iterated logarithm. Further applications for random variables in the domain of attraction of a stable law are presented.