Precise asymptotics for record times and the associated counting process. (Q2574510): Difference between revisions

Let \( X_1, X_2,\dots , \) be i.i.d.\ absolutely continuous random variables, \(L_n\) be their record times, i.e.\ \(L_1=1,\, L_n = \min \{k: X_k > X_{L(n-1)}\},\, n\geq 2, \) and \(\mu (n)\) be the associated counting process of records, \(\mu (n) = \max \{k: L(k) \leq n\}.\) A precise asymptotics is proved for infinite sums of type \(\sum \frac {(\log n)^\delta }{n} P(| \mu (n) - \log n | > \varepsilon (\log n)^{1/p})\), respectively of type \(\sum \frac {(\log n)^\delta }{n} P(| \log L(n) - n| > \varepsilon (\log n)^{1/p} )\) with properly chosen \(\delta \) and \(p\). In the proofs the author's previous results on convergence rates for record times and associated counting processes are utilized.