The order of magnitude of the moments of the modulus of continuity of multiparameter Poisson and empirical processes (Q1086934): Difference between revisions

Let \(N_ n(t)\) be a d-dimensional Poisson process with \(E(N_ n(t))=n F(t)\). Then the standardized Poisson process, restricted to the unit cube is \[ Z_ n=\{Z_ n(t)=n^{-1/2}(N_ n(t)-n F(t)),\quad t\in [0,1]^ d\}. \] It is shown that for \(\delta\in (0;0.5)\) there exist \(0<C_ 1,C_ 2<\infty\) so that for all \(n\geq m^{1+2\delta}:\) \[ C_ 1\leq (m/\ln (m))^{\nu /2} E((\max_{R\in P_ m}\sup_{s,t\in R}| Z_ n(s)-Z_ n(t)|)^{\nu})\leq C_ 2 \] where \(P_ m\) is the partition of the unit cube into squares of equal size. Because the law of the process \(Z_ n\), conditional on \(N_ n(1,...,1)=n\), equals that of the multivariate empirical process, the result is seen to hold for the latter, too.