Extremal stochastic integrals: a parallel between max-stable processes and \(\alpha\)-stable processes (Q2463680)

The paper is devoted to construction of extremal stochastic integrals by random \(\alpha\)-Fréchet sup-measures and investigation of their properties, specially, connections with \(\alpha\)-stable integrals. A r.v. \(\xi\) has \(\alpha\)-Fréchet distribution \(F(\alpha,\sigma)\) if \(P\{\xi\leq x\}=\exp(-\sigma^\alpha x^{-\alpha})\), \(x>0\). A random set function \(M(A)\), \(A\subseteq R\) is an \(\alpha\)-Fréchet sup-measure with a (nonrandom) control measure \(\mu\) if for any disjoint Borel \(A_1,A_2\in R\), \(M(A_1\cup A_2)= \max(M(A_1),M(A_2))\) and \(M(A_i)\) are independent \(F(\alpha,\mu(A_i)^{1/\alpha})\). The extremal integral for a simple function is defined as a usual integral but with sum replaced by max. It is extended by continuity on arbitrary positive functions. Connections of extremal integrals with de Haan's spectral representation through the LePage type series representation of max-stable processes is obtained. Necessary and sufficient conditions for a max-stable \(\alpha\)-Fréchet process to have an extremal integral representation are derived. Examples of classes of max-stable processes such as moving maxima processes are considered.