Small deviation estimates for some additive processes

From MaRDI portal





Let \(X_1,\ldots,X_d\) be independent, real-valued stochastic processes over some index set \(T\). Then one may define the generated additive process \(\mathbb X=(\mathbb X(t))_{t\in T^d}\) by \[ \mathbb X(t) :=\sum_{j=1}^d X_j(t_j),\quad t=(t_1,\dots,t_d)\;. \] Now suppose that there are some \(\beta_j>0\) such that \[ \lim_{\varepsilon\to 0} \varepsilon^{\beta_j}\,\log\mathbb{P}\left(\sup_{t\in T}| X_j(t)| \leq \varepsilon\right) := -A_j \] exists in \((0,\infty)\) for \(1\leq j\leq d\). Under some natural assumption, satisfied by the most important classes of random processes, as e.g.~by Gaussian ones, this implies \[ \lim_{\varepsilon\to 0} \varepsilon^{\beta}\,\log\mathbb{P}\left(\sup_{t\in T}| \mathbb X_j(t)| \leq \varepsilon\right) = -\left(\sum_{j=1}^m A_j^{1/(1+\beta)}\right)^{1+\beta} \] where \(\beta=\max_{j\leq d}\beta_j\) and \(m\leq d\) denotes the multiplicity of the largest \(\beta_j\). As an application, limit inf type LIL are proved for additive stable processes. For the entire collection see [Zbl 1029.00032].











This page was built for publication: Small deviation estimates for some additive processes

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4473013)