Stationary systems of Gaussian processes (Q614126): Difference between revisions

A Poisson point process on \(\mathbb{R}\) with an intensity measure \(m\) is considered. To every point of the process \(U_i\) corresponds independently a Gaussian process (field) \(\xi_i(t)\) \((t\in \mathbb{R}^n,\) \({n\geq1)},\) having identical distribution for different \(i\). The system of processes \(V_i(t)=U_i+\xi_i(t)\) is investigated. This system is proved to be invariant relative to a shift (stationary) if and only if its representative \((m,\xi)\) belongs to one of three classes: (1) \(m\) is arbitrary and \(\xi\) is a stationary Gaussian process; (2) the Poisson process is homogeneous on the line, and \(\xi \) is a Gaussian process with zero mean and stationary increments \(W(t)\) plus some additive process; (3) the density of the measure \(m\) is equal to \(\alpha e^{-\lambda x}\) \((x\in \mathbb{R}, \,\alpha>0,\,\lambda\in \mathbb{R})\), and \(\xi(t)=W(t)-\lambda\sigma^2(t)/2+c\), where \(\sigma^2(t)\) is the variance of \(W(t)\), and \(c\in \mathbb{R}\).