On the law of the iterated logarithm for trigonometric series with bounded gaps II
arXiv:1403.1629
Abstract
It is well-known that for a quickly increasing sequence the functions show a behavior which is typical for sequences of independent random variables. If the growth condition on is relaxed then this almost-independent behavior generally fails. Still, probabilistic constructions show that for \emph{some} very slowly increasing sequences this almost-independence property is preserved. For example, there exists having bounded gaps such that the normalized sums satisfy the central limit theorem (CLT). However, due to a ``loss of mass'' phenomenon the variance in the CLT for a sequence with bounded gaps is always smaller than . In the case of the law of the iterated logarithm (LIL) the situation is different; as we proved in an earlier paper, there exists with bounded gaps such that In the present paper we prove a complementary results showing that any prescribed limsup-behavior in the LIL is possible for sequences with bounded gaps. More precisely, we show that for any real number there exists a sequence of integers satisfying such that the limsup in the LIL equals for almost all . Similar results are proved for sums and for the discrepancy of .
This manuscript is a complement to the paper "On the law of the iterated logarithm for trigonometric series with bounded gaps", Probab. Th. Rel. Fields, 154 (2012), no. 3-4, 607--620, by the same authors