Erdős-Rényi-type laws applied to Gaussian processes (Q757977)

Let \(\{\xi\) (t): \(t\geq 0\}\) be a continuous centered Gaussian process with \(\xi (0)=0\) and stationary increments \(E(\xi (t)-\xi (s))^ 2=\sigma^ 2(| t-s|).\) The author extends some of the results of \textit{M. Csörgö} and \textit{P. Révész} [Ann. Probab. 7, 731-737 (1979; Zbl 0412.60038)] on large increments of a Wiener process to \(\{\xi\) (t)\(\}\) provided \(\sigma\) (t) satisfies a number of regularity assumptions, e.g. \(\sigma\) (t) is assumed to be regularly varying with index \(0<\gamma <1\). Moreover, analogous results are proved for stationary Gaussian processes under certain regularity conditions on the covariance function \(r(t)=E \xi (0)\xi (t)\), like r(t) being convex or lim \(t^{1+\nu}[dr(t)/dt]=0\) for some \(\nu >0\). The increment function \(\{a_ T:\) \(T>0\}\) and the function r(t) are definitely related, the stronger the assumptions on r(t), the larger increments can be characterized.