Asymptotic distribution of the maximum of n independent stochastic processes

continuous in probability functional limit theorem max-infinitely divisible max-self-decomposable maxima Poisson process triangular array of processes

Mathematics Subject Classification ID

Central limit and other weak theorems (60F05) Functional limit theorems; invariance principles (60F17) Extreme value theory; extremal stochastic processes (60G70) Order statistics; empirical distribution functions (62G30)

Review text

A stochastic process \(X=\{X(t),t\in T\}\) is max-infinitely divisible if for every \(r\) there are i.i.d. processes \(Z_ i=\{Z_ i^{(r)}(t),t\in T\}\) such that \(X\) has the same distribution as \(\max\{Z_ 1,\dots,Z_ r\}\), the max being taken pointwise. Equivalently: \(X\) is the limit in distribution of the row maxima in a triangular array of processes, i.i.d. in each row. The process \(X\) is max-self-decomposable if it is the limit in distribution of normed and centered partial maxima of a sequence of independent, not necessarily i.i.d., processes. The authors characterize these limiting processes for \(T=\mathbb{N}\) and for \(T=\mathbb{R}\) when continuous in probability. Let \(\{X_{jn}(t),t\in\mathbb{R}\}\), \(j=1,\dots,n\), for fixed \(n\) be i.i.d. and let the process \(\{X_ n(t),t\in\mathbb{R}\}\), with \(\{X_ n(t)=\max\{X_{jn}(t),j=1,\dots,n\}\}\), converge in distribution as \(n\to\infty\), with respect to finite-dimensional marginals, to a right- continuous (in probability) process \(\{X'(t),t\in\mathbb{R}\}\). Then \(X'\) has the same distribution as \(X\) with \(X(t)=\sup\{f_ t(Y_ k),k\in\mathbb{N}\}\) where \(Y_ 1,Y_ 2,\dots\) is an enumeration of the points of a homogeneous Poisson process on \(\mathbb{R}\) and \(f_ t\), \(t\in\mathbb{R}\), is a family of measurable functions \(f_ t:\mathbb{R}\to[-\infty,\infty)\) such that \(\lambda\{s\mid f_ t(s)>c_ t\}<\infty\) for some \(c_ t\in\mathbb{R}\), with \(\lambda\) Lebesgue measure on \(\mathbb{R}\). This result is deduced from the corresponding one for \(T=\mathbb{N}\), where also the joint distribution function of the \(X(k)\) is derived. Let \(\{X_ n(t),t\in\mathbb{R}\}\), \(n=1,2,\dots\), be independent processes and let the process \(\max\{a_ n^{-1}(X_ k(t)-b_ n),\;k=1,\dots,n\}\) converge in distribution, as \(n\to\infty\) to a nondegenerate continuous (in probability) process \(\{X(t),\;t\in\mathbb{R}\}\). Under the usual conditions on \(a_ n>0\) and \(b_ n\), the self-max-decomposable process \(X\) may be represented by means of a homogeneous Poisson process on \(\mathbb{R}^ 2_ +\). The proof starts with the corresponding theorem for \(T=\mathbb{N}\) and then also gives the joint d.f. of the \(X(k)\). The paper also gives a calculus and convergence theorems for random variables of the above type \(\sup f(Y_ k)\) for a possible nonhomogeneous Poisson process.

Recommendations

Cited in

(21)

This page was built for publication: Asymptotic distribution of the maximum of n independent stochastic processes

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5285997)