Extrema of some Gaussian processes with large trends and density estimation in \(L_{\infty}\)-norm (Q1120930): Difference between revisions

Let \(\{\mu_ n(t)\), \(t\in [0,T_ n]\}\) be a sequence of trends and \(X_ n-\mu_ n(t)\) a stationary Gaussian process having mean zero. This paper establishes the exact asymptotic behaviour for the probabilities \[ \Pr (\max_{[0,T_ n]}| X_ n(t)| \leq u_ n),\quad \lim_{n\to \infty}T_ n=+\infty, \] when the \(\{\mu_ n(t)\}\) are not asymptotically negligible and essentially contribute to the approximating distribution. The results are used to study the optimal rates of convergence for nonparametric density estimation when the closeness is measured in \(L_{\infty}\) norm.