Sample and ergodic properties of some min-stable processes (Q1176153): Difference between revisions

A min-stable random vector \((X_ 1,\dots,X_ k)\) is characterized by the fact that for all \(a_ 1 \geq 0,\dots,\) \(a_ k \geq 0\) with \(\sum^ n_ 1 a_ i > 0\) the random variable \(\bigwedge^ k_ 1 X_ i/a_ i\) follows an exponential distribution. A min-stable process \(X(t)\) with either continuous or discrete time is a process such that all the corresponding finite-dimensional distributions are min-stable. Such processes may be conveniently defined by the representation \(X(t) = \bigwedge^ \infty_{l = 1} Y_ l/f_ t(U_ l)\), where \((U_ l,Y_ l)\), \(l = 1,2,\dots,\) define a unit-intensity two-dimensional Poisson process on \([0,1] \times \mathbb{R}^ +\) and the \(f_ t:[0,1] \to \mathbb{R}^ +\) are in \(L_ 1(0,1)\). A stationary min-stable process is obtained when the spectral functions \(f_ t\) are of the form \(f_ t = \Gamma^ tf\), where \(\Gamma^ t\) is a power group of unitary operators called pistons. An interesting subclass of min-stable processes is the moving minimum process introduced by the reviewer [J. Multivariate Anal. 13, No. 2, 257- 272 (1983; Zbl 0519.60045) and Multivariate analysis, Proc. 6th Int. Symp., Pittsburgh/Pa. 1983, Multivariate Anal. 6, 145-164 (1985; Zbl 0606.60047)]. In this case, \(f_ t(u)\) is of the form \(f(u-t)\). The author proves a series of results concerning such moving minimum min- stable processes. Define the association between \(X_ s\) and \(X_ t\) as the probability that \(X_ s = Y_ l/f_ s(U_ l)\) and \(X_ t = Y_ l/f_ t(U_ l)\) for the same \(l\). Among other results, the author shows that the association \(q_ 0(s)\) between \(X(0)\) and \(X(s)\) tends to 0 as \(s\to \infty\) for all moving minimum processes (0-mixing). Other types of mixing-type conditions are characterized in terms of the behavior of \(f\) at infinity.