On the nonexistence of a law of the iterated logarithm for weighted sums of identically distributed random variables (Q751708): Difference between revisions

Let \(\{X,X_ n\), \(n\geq 1\}\) be a sequence of i.i.d. random variables whose common distribution is of the form \(P\{| X| >x\}=L(x)x^{- \alpha}\) for \(x\geq 1\), and \(=1\) elsewhere, where L(x) is slowly varying at infinity. Let \(\{a_ n\), \(n\geq 1\}\) be given finite real constants and let \(S_ n=\sum^{n}_{k=1}a_ kX_ k\), \(n\geq 1\). Assume \(\alpha\in [0,1)\cup (1,2).\) A first result shows that if \(\Sigma =\sum^{\infty}_{n=1}P\{| X| >b_ n/| a_ n| \}=\infty\) for some \(\{b_ n\), \(n\geq 1\}\) satisfying \(b_ n=O(b_{n+1})\), \(b_ n\to \infty\), then \(\limsup_{n\to \infty}S_ n/b_ n=\infty\) a.s.. Next, if \(0<b_ n\uparrow \infty\) and \(\Sigma <\infty\), then it is proven that \(S_ n/b_ n\to 0\) a.s.. As a consequence a law of the iterated logarithm for \(S_ n\) does not exist.