Precise asymptotics for a type of order statistics (Q2463685)

Order statistics \(X_{N(t),N(t)}\leq X_{N(t)-1,N(t)}\leq\dots,\leq X_{1,N(t)}\) are considered for an i.i.d. sample of random size \(N(t)\), where \(N(t)\) is a counting process. The authors investigate the convergence of sums \[ G_k(\varepsilon)\sum_{n=n_0}^\infty \varphi(n) P\{X_{k,N(n)}\geq\varepsilon c_nh(n)\}=1 \] as \(\varepsilon\to 0\) for different normalizing sequences \(G_k\), \(\varphi\), \(c_n\) and \(h\) in the case where \(X_i\) belong to the domain of max-attraction of the Fréchet distribution.