Sample path properties of ergodic self-similar processes (Q2640226)

Let \(X=\{X(t)\), \(t\geq 0\}\) be a self-similar process with exponent \(\kappa >0\) such that \(X(0)=0\). Assuming every shift transformation of the stationary process Y defined by \(Y(t)=e^{-\kappa t}X(e^ t)\), \(- \infty <t<\infty\), is ergodic, the author gives several zero-one laws on the growth properties of sample paths of X. For example he proves that the probability of the existence of \(\delta >0\) with \(| X(t)| \leq g(t)\) for \(0<t<\delta\) is zero or one, where \(g(t)=t^{\kappa}\phi (t)\) with positive increasing \(\phi\). Similar results are obtained for local growth at \(\infty\) and uniform growth. He obtains also zero-one laws for several other sample path properties including Hausdorff dimension of the graph of X. He next considers a class of stable self-similar processes with stationary increments which is represented by the integral with respect to a strict stable process, and gives integral criteria for a function to belong to the upper and lower class with respect to the local growth. Similar criteria are given for uniform growth of fractional stable processes. Three exponents are defined for a self-similar process X. The first is the exponent of local growth of sample paths, the second is that of uniform growth and the last is the exponent of self- similarity. The author shows that these three are equal to one another if the distribution of X(t) has exponential tail.