An almost sure central limit theorem for self-normalized products of sums of i.i.d. random variables (Q624552): Difference between revisions

Let \((X_n)\) be a sequence of i.i.d. positive random variables with mean \(\mu\) and let \(S_n= X_1+\cdots+ X_n\). If \(X_1\) has finite variance \(\sigma^2\) then Gonchigdanzan and Rempala proved the following almost sure central limit theorem (ASCLT) \[ (\log n)^{-1} \sum^n_{k=1} {1\over k}\mathbf{1}\Biggl\{\prod^k_{j=1} \Biggl({S_j\over j\mu}\Biggr)^{\mu/\sigma\sqrt{k}}\leq x\Biggr\}\to \Phi((\log x)/\sqrt{2})\text{ a.s.} \] for any \(x> 0\), where \(\Phi(x)\) is the standard normal distribution function. In the present paper it is shown that replacing \(\sigma\sqrt{k}\) in the exponent by \((\sum^k_{j=1} (X_j- \mu)^2)^{1/2}\) or by \((\sum^k_{j=1} (X_j-\overline X_k)^2)^{1/2}\) (i.e. using a self-normalization) the above ASCLT holds under a weaker assumption that the distribution of \(X_1\) belongs to the domain of attraction of the normal law.