A general law of precise asymptotics for the counting process of record times (Q1414232)

Let \(\{X_{n}\), \(n\geq 1\}\) be i.i.d. random variables with partial sums \(\{S_{n}\), \(n\geq 1\},\) and assume that the distribution function of \(X_{1}\) is absolutely continuous. Put \(L(1)=1,\) define recursively the record times \(L(n)=\min \{k:X_{k}>X_{L(n-1)}\}\), \(n\geq 2,\) and consider the associated counting process \(\{\mu (n)\), \(n\geq 1\},\) where \(\mu (n)=\max \{k:L(k)\leq n\}.\) \textit{A. Gut} [Stochastic Processes Appl. 101, 233-239 (2002)] obtained four precise asymptotics concerning the counting process such as the next one. Theorem. For \(r>0,\) \[ \lim_{\varepsilon \searrow \sqrt{2r}}\sqrt{\varepsilon ^{2}-2r}\sum_{n\geq 9}(\log \log n)^{r-1}(n\log n)^{-1}P(\left|\mu (n)-\log n\right|>\varepsilon \sqrt{\log n\log \log \log n})=\sqrt{2/r}. \] The authors consider series of the form \[ f(\varepsilon)=\sum_{n\geq n_{0}}\varphi (n)P(\left|\mu (n)-E\mu (n)\right|>\varepsilon \sqrt{\log n} h(n)),\quad \varepsilon >0, \] where \(g\) and \(h\) are positive, strictly increasing and differentiable functions on \([n_{0},\infty)\) with \(\lim_{x\rightarrow \infty }g(x)=\lim_{x\rightarrow \infty }h(x)=\infty ,\) and \(\varphi =(g\circ h)^{\prime }.\) Under some regularity conditions on \(g\) and \(h,\) they claim they find a threshold \(a\) and a normalizing function \(G(\varepsilon)>0, \varepsilon >a,\) with \(\lim_{x\searrow a}G(\varepsilon)=\infty \) such that \(\lim_{x\searrow a}G(\varepsilon)^{-1}f(\varepsilon)=1.\) Unfortunately, the paper is poorly and disorderly written, so that it is hard to realize whether the goal has been achieved. The wrong English makes even more difficult the reading. Here are two examples: ``\(S_{n}=\sum_{i=1}^{k}X_{k}, n\geq 1\)'';``Another extension departs from the observation that the convergence rate and limit value of \(\sum_{n=1}^{\infty }\varphi (n)P(|S_{n}|\geq \varepsilon f(n))\) as \(\varepsilon \downarrow a, a\geq 0\).''