Hausdorff measure of the sample paths of Gaussian random fields (Q1355279): Difference between revisions

Let \(Y(t)\) \((t\in \mathbb{R}^N)\) be a real-valued, centered Gaussian random field with stationary increments and \(Y(0)=0\). Assume that there exist positive constants \(\delta_0\), \(c_1\) and \(c_2\) such that for any \(t\in\mathbb{R}^N\) and \(h\in\mathbb{R}^N\) with \(|h|\leq \delta_0 \) \[ E\biggl[ \bigl(Y(t+h) -Y(t)\bigr)^2 \biggr]\leq c_1 \sigma^2 \bigl(|h|\bigr) \] and for any \(0<r\leq \min\{|t|, \delta_0\}\) \[ \text{Var} (Y(t) \mid Y(s): r\leq|s-t|\leq\delta_0) \geq c_2 \sigma^2(r), \] where \(\sigma: [0,\delta_0) \to[0,\infty)\) is a nondecreasing function which is regularly varying at the origin with index \(\alpha\) \((0<\alpha<1)\). Let \(\psi\) be the inverse function of \(\sigma\). Consider the \((N,d)\) Gaussian random field defined by \(X(t)= (X_1(t), \dots, X_d(t))\) \((t\in \mathbb{R}^N)\), where \(X_1, \dots, X_d\) are independent copies of \(Y\). It is shown that if \(N<\alpha d\), then \(0<\varphi\text{-m}(X([0,1]^N)) <\infty\) a.s., where \(\varphi(s) =\psi (s)^N \log\log {1\over 2}\) and \(\varphi\)-m denotes the \(\varphi\)-Hausdorff measure. This generalizes the results of \textit{A. Goldman} [``Mouvement brownien à plusieurs paramètres: mesure de Hausdorff des trajectoires'' (1988; Zbl 0681.60040)] and \textit{M. Talagrand} [Ann. Probab. 23, No. 2, 767-775 (1995; Zbl 0830.60034)] for fractional Brownian motion. Further results on the graph, level sets of Gaussian random fields have also been proved by the author in two papers [Math. Proc. Camb. Philos. Soc. 120, No. 3, 535-546 (1996; Zbl 0864.28001) and Probab. Theory Relat. Fields (to appear)].