Large deviations for the maxima of some random fields (Q1088280)

For the approximation of the probabilities of large deviations of Kolmogorov-Smirnov statistics in the two-dimensional case and the same probabilities for test statistics connected with the change point problem the asymptotical behaviour of probabilities of large deviations for the maxima of some random fields (closely related to simple one-dimensional processes such as random walk, Brownian motion and Brownian bridge) is studied. In particular, for the standard Brownian motion \(W(t)\), \(0\leq t<\infty\), it is proved that if \(\mu >0\), m is a positive integer and \(m,u\to \infty\) such that \(m\mu u^{-1} = \text{ const}\in (1,\infty)\), then \[ P\{\max_{s\leq t\leq m}[W(t)-W(s)-\mu (t-s)]>u\}=[2\mu (m\mu - u)+3+o(1)]\exp (-2\mu u). \]