Sampling at a random time with a heavy-tailed distribution (Q2708441)

Let \(\{X(t)\}\) be a renewal process with epochs \(S_n\) and \(T\) an independent heavy-tailed random variable. The asymptotics of \(P(X(T)> n)\), \(n\to\infty\), and more generally of \(Ee^{-g(S_n)}\) is studied as \(n\to\infty\), thereby extending \textit{S. Asmussen}, \textit{C. Klüppelberg} and \textit{K. Sigman} [Stochastic Processes Appl. 79, No. 2, 265-286 (1999; Zbl 0961.60080)] who dealt with the Poisson case. As in the quoted paper two different scenarios emerge according to whether \(\log P(T> t)\) decreases slower (as in the regularly varying case) or faster than \(-\sqrt t\). Applications to the GI/G/1 queue are given. The proofs rely heavily upon large deviations techniques.