Discrete and continuous time extremes of Gaussian processes (Q2488450): Difference between revisions

Let \(X_t\) be a zero mean Gaussian stationary process with covariance \(r(t)=1-| t| ^\alpha+o(| t| ^\alpha)\) (as \(t\to\infty\), \(\alpha\in(0,2)\)), let \({\mathcal R}(\delta)=\{k\delta,\;k=0,1,2,\dots\}\) be a uniform grid on \(R_{+}\), and let \(M_T=\max_{t\in[0,T]} X_t\), \(M_T^\delta=\max_{t\in {\mathcal R}(\delta)\cap[0,T]} X_t\). The author investigates the limits \[ P_\delta=\lim_{T\to\infty}{\mathbf P}(a_T(M_T-b_T)<x,\;a_T(M_T^\delta-b_T^\delta)<y) \] for suitable normalizing constants and different choices of \(\delta\). It is shown that \(P_\delta=\exp(-e^{-x}-e^{-y})\) for sparse grids (\(\delta=\delta(T)=o((\log T)^{-1/\alpha})\)), and \(P_\delta=\exp(-e^{-x\wedge y})\) for dense grids (\(\delta(T)(\log T)^{1/\alpha}\to\infty\)). The limit distribution in the intermediate case of Pickand's grid is also obtained.