On limiting cluster size distributions for processes of exceedances for stationary sequences (Q613169)

Let \(\{X_{k}\}_{k\geq 0}\) be a stationary discrete time real-valued process. \ For a suitably chosen increasing real sequence \(\{u_{n}\}\), consider the point process of exceedances \[ N_{n}(A)=\#\{k/n\in A:X_{k}>u_{n}\},\quad A\in \mathcal{B}(\mathbb{R}_{+}). \] Under broad assumptions, if the process \(N_{n}\) has a limit as \(n\rightarrow \infty \), the latter must be a compound Poisson process, with some compounding law \(G=\{g_{k}\}_{k\geq 1}\) on \(\mathbb{N}\). Set \(\mu =\sum_{k=1}^{\infty }kg_{k}\). The authors show that, for any given distribution \(G\) on \(\mathbb{N}\) with \( \mu \leq \infty \), there exists a stationary sequence for which \(G\) is the compouding law for the limiting point process of exceedances.