scientific article; zbMATH DE number 1409911

From MaRDI portal
Publication:4940332





Let \(\varphi(x)\) be an Orlicz \(N\)-function and let \(f(x)\) be an Orlicz \(S\)-function. Let \(\xi(t)\), \(t\in T,\) be a stochastic process such that the Cramér condition holds true \((E\exp\{\lambda\xi(t)\}\leq C<\infty, t\in T, |\lambda|<\lambda_0)\), and let \(\|\cdot\|\) be a norm on the space of random variables such that \(\forall t,s\in T\) the following inequalities hold: \[ E\exp(\lambda\xi(t))\leq\exp\{\varphi(\lambda\|\xi(t)\|)\},\qquad 0<\lambda\leq 1/f(\|\xi(t)\|); \] \[ E\exp(\lambda(\xi(t)-\xi(s)))\leq\exp\{\varphi(\lambda\|\xi(t)-\xi(s)\|)\},\qquad 0<\lambda\leq{1/f(\|\xi(t)-\xi(s)\|)}. \] It is proved that \(P\{\sup_{t\in T}\xi(t)>x\}\leq U(x)\) and \(P\{\sup_{t\in T}\|\xi(t)\|>x\}\leq 2U(x)\), where \(U(x)\) is described in terms of the metric entropy \(H(u)\) by the norm \(\|\cdot\|\). The entropy Dudley integral \(\int_0^p H(u)/\varphi^{(-1)}(H(u)) du\) plays the key role in proving the inequalities.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4940332)