Chover's law of the iterated logarithm and weak convergence (Q760708): Difference between revisions

Let \(X_ 1,X_ 2,..\). be a sequence of i.i.d. random variables defined over a common probability space (\(\Omega\),\({\mathcal B},P)\) and let \(S_ n=X_ 1+X_ 2+...+X_ n\), \(n\geq 1\). Assume that there exists a sequence \((B_ n)\) of positive constants \((B_ n\to \infty\) as \(n\to \infty)\) such that for all \(x\in (-\infty,\infty)\) \(\lim_{n\to \infty}P(S_ n\leq xB_ n)=G_{\alpha}(x)\), where \(G_{\alpha}\) is a stable distribution function with characteristic exponent \(\alpha\), \(0<\alpha \leq 2.\) In this paper the following law of the iterated logarithm is obtained - \[ \limsup_{n\to \infty}| B_ n^{-1}S_ n|^{1/(\log \log n)}=\theta \quad a.s., \] where \(\theta =e^{1/\alpha}\) whenever \(\alpha\in (0,2)\) and \(\theta \in [1,e^{1/2}]\) whenever \(\alpha =2\). When \(\alpha =2\), it is further shown that for each \(\theta \in [1,e^{1/2}]\), there corresponds a sequence \((X_ n)\) of i.i.d. random variables such that the above law of the iterated logarithm holds. Also, it is shown through an example that for the above law of the iterated logarithm to hold it is not necessary that the sequence \((S_ n)\), properly normalised, should converge weakly.