The finite mean LIL bounds are sharp (Q797898): Difference between revisions

The following problem is discussed in this paper: suppose \(X_ 1,X_ 2,X_ 3,..\). are i.i.d., \(S_ n=\sum^{n}_{j=1}X_ j\), \(EX_ 1=0\), how can we put \(\{a_ n\}\), such that \(L=\lim \sup_{n\to\infty }S_ n/a_ n<\infty ?\) A result in this direction is the famous law of iterated logarithm where \(L=1\). The author proved [see also Z. Wahrscheinlichkeitstheor. Verw. Geb. 39, 151-165 (1977; Zbl 0387.60033)], that if we put \(a_ n=(\log \log n)K(n/\log \log n)\) with K(y) satisfying \(yE\{| X_ 1/K(y)|^ 2\wedge| X_ 1/K(y)|\}=1\) (for \(y>0)\), then 1\(\leq L\leq 1.5\) iff \(P\{X_ n>a_ n\), \(i.o.\}=0\). An example for which \(L=1.5\), is given.