Large deviations for stationary Gaussian processes (Q1103949)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Large deviations for stationary Gaussian processes
scientific article

    Statements

    Large deviations for stationary Gaussian processes (English)
    0 references
    0 references
    1985
    0 references
    The authors derive a so-called level 3 large deviation principle for stationary ergodic Gaussian sequences \(X_ k\), \(k\in {\mathbb{Z}}:\) For \(\omega =\{x_ k\}\) in the path space \({\mathbb{R}}^{{\mathbb{Z}}}\), let \(\omega^{(n)}\) be the sequence made periodic with period n and agreeing with \(\omega\) at coordinates 1 to n. Then \(\pi_ n(\omega)\) is defined by \[ \pi_ n(\omega)=n^{-1}\sum^{n-1}_{j=0}\delta_{T\quad j\omega}(n) \] where T j is the usual shift. Then for a Borel subset A in the space of stationary measures on the path space \[ -\inf_{R\in int(A)}H(R)\leq \overline{\lim}\overline{_{n\to \infty}}n^{-1} \log P(\pi \quad_ n\in A)\leq -\inf_{R\in cl(A)}H(R) \] where inf and cl refer to the weak topology. The entropy function H can be given explicitly in terms of the spectral density.
    0 references
    large deviation principle for stationary ergodic Gaussian sequences
    0 references
    stationary measures on the path space
    0 references
    entropy function
    0 references
    spectral density
    0 references

    Identifiers