A note of the behavior of sample statistics when the population mean is infinite (Q801379)

Let \(X_ i\geq 0\) be i.i.d. random variables with \(EX_ i=\infty\). The authors give conditions (Theorem 1) for a real function \(\phi\) under which \(\overline{\phi (X)}/\phi (\bar X)\to^{a.s.}0\). As applications of this theorem they obtain asymptotic behaviour of some sample statistics when the population mean is infinite. Among other things they have: 1. If \(X_ 1,X_ 2,..\). are i.i.d., \(X_ i\geq 0\), \(EX_ 1=\infty\), and \(\phi (x)=x^{\rho}L(x)\) for some \(0<\rho <1\) and slowly varying function L then \[ n^{-1}\sum^{n}_{i=1}\phi (X_ i)/\phi (n^{- 1}\sum^{n}_{i=1}X_ i)\to^{a.s.}0,\quad n\to \infty. \] 2. Let \(X_ 1,X_ 2,..\). be i.i.d. random variables. If \(EX^ 2_ 1=\infty\) then \[ \left( \begin{matrix} n\\ 2\end{matrix} \right)^{-1}\sum_{1\leq i\leq j\leq n}X_ iX_ j/(1/n)\sum^{n}_{i=1}\quad X^ 2_ i\to^{a.s.}0,\quad n\to \infty. \] 3. Let \(X_ 1,X_ 2,..\). be i.i.d. with \(E| X_ 1|^ p=\infty\) for some \(p>1\). Then \[ \sum^{n}_{i=1}| X_ i-\bar X|^ p/\sum^{n}_{i=1}| X_ i|^ p\to^{a.s.}1,\quad n\to \infty,\quad \bar X=n^{- 1}\sum^{n}_{i=1}X_ i. \] 4. Let \(X_ 1,X_ 2,..\). be i.i.d., \(X_ 1\geq 0\). Then a necessary and sufficient condition for \[ (X_ 1X_ 2...X_ n)^{1/n}/n^{-1}\sum^{n}_{i=1}X_ i\to^{a.s.}0,\quad n\to \infty \] is \(E(X_ 1-\log X_ 1)=\infty\).