Persistence and Ball Exponents for Gaussian Stationary Processes

TL;DR

This paper investigates the persistence and ball exponents of Gaussian stationary processes, establishing conditions for their existence, positivity, and continuity with respect to spectral measures and parameters.

Contribution

It provides new results on the existence, positivity, and continuity of persistence and ball exponents for Gaussian stationary processes based on spectral measure properties.

Findings

Persistence exponent exists if spectral density is positive at zero.

Positivity of the exponent is equivalent to having an absolutely continuous spectral component.

Continuity of exponents in parameters and spectral measure is established.

Abstract

Consider a real Gaussian stationary process $f_\rho$, indexed on either $\mathbb{R}$ or $\mathbb{Z}$ and admitting a spectral measure $\rho$. We study $\theta_{\rho}^\ell=-\lim\limits_{T\to\infty}\frac{1}{T} \log\mathbb{P}\left(\inf_{t\in[0,T]}f_{\rho}(t)>\ell\right)$, the persistence exponent of $f_\rho$. We show that, if $\rho$ has a positive density at the origin, then the persistence exponent exists; moreover, if $\rho$ has an absolutely continuous component, then $\theta_{\rho}^\ell>0$ if and only if this spectral density at the origin is finite. We further establish continuity of $\theta_{\rho}^\ell$ in $\ell$, in $\rho$ (under a suitable metric) and, if $\rho$ is compactly supported, also in dense sampling. Analogous continuity properties are shown for $\psi_{\rho}^\ell=-\lim\limits_{T\to\infty}\frac{1}{T} \log\mathbb{P}\left(\inf_{t\in[0,T]}|f_{\rho}(t)|\le \ell\right)$, the ball exponent of $f_\rho$, and it is shown to be positive if and only if $\rho$ has an absolutely continuous component.