Point processes associated with stationary stable processes (Q2485804)

Let \(\{X_n\}\) be a stationary symmetric \(\alpha\)-stable (\(\alpha\in (0,2)\)) process, i.e. the linear combinations \(\sum_{n=1}^k c_n X_n\) have, for all choices of \(k\) and real numbers \(c_1\),\dots,\(c_k\), a symmetric \(\alpha\)-stable distribution whose characteristic function is given by \(\varphi(\theta) =\exp\{-\sigma^{\alpha}| \theta| ^{\alpha}\}\). The scaling constant \(\sigma\) depends on \(k\) and the choice of \(c_1, \dots ,c_k\). The authors follow the notation of \textit{G. Samorodnitsky} and \textit{M. Taqqu} [``Stable non-Gaussian random processes'' (1994; Zbl 0925.60027)]. For a sequence of positive constants \(b_n\uparrow \infty\), consider a sequence of point processes \(N_n=\sum_{k=0}^{n-1} \delta_{b_n^{-1}X_k}\), \(n=1,2,\dots\), on \([-\infty, \infty]\setminus\{0\}\), \(\delta_x\) is the point mass at \(x\). Weak convergence in the space \({\mathcal M}\) of Radon measures on \([-\infty, \infty]\setminus\{0\}\) of sequences of point processes \(N_n\) is handled by extreme value techniques, see \textit{S. Resnik} [``Extreme values, regular variation, and point processes'' (1987; Zbl 0633.60001)]. For an independent and identically \(\alpha\)-stable distributed sequence \(\{X_k\}\) satisfying the condition \(P(| X_k| >\lambda) \sim C_{\alpha} \sigma^{\alpha}\lambda^{-\alpha}\) as \(\lambda \to \infty\), \(C_{\alpha}\in(0, \infty)\) (see Samorodnitsky and Taqqu, loc. cit.), it is well known that an acceptable choice of the scaling sequence is \(b_n=n^{1/\alpha}\) and in this case the sequence \(N_n\) converges weakly in the space \({\mathcal M}\) (with the vague topology) to a very particular Poisson random measure, whose intensity blows up near the origin (see Resnik, loc. cit.) It is natural that attention has been focused on removing the assumption of independence in the original process \(\{X_k\}\). The general sense of the obtained results was that if \(\{X_k\}\) is a stationary process with sufficiently weak dependence, then the sequence \(N_n\) still converges weakly, and with the same sequence of normalizing constants \(b_n=n^{1/\alpha}\), the limiting random measure is, typically, a cluster Poisson process, see \textit{R. Davis} and \textit{T. Hsing} [Ann. Probab. 23, 879--917 (1995; Zbl 0837.60017)]. The aim of this paper is to understand what may happen when the dependence in the process \(\{X_k\}\) is no longer weak or local. The article is organized as follows. In Section 2, background information and set up of the framework of our study are given. In Section 3 point processes corresponding to dissipative maps are studied; these turn out to be processes based on mixed moving averages. Section 4 considers the more intricate case where the stationary stable process is associated with a conservative map.