Local and integral theorems on large deviations of the maximum point of a sum of random fields (Q1920108): Difference between revisions

Let \(C(K)\) denote the usual Banach space of real-valued continuous functions on a compact set \(K\subset R^k\). Let \(\xi_i\), \(i\geq 1\), be i.i.d. random elements with values in \(C(K)\) and let \(S_n(t)=\xi_1(t)+\cdots+ \xi_n(t)\). Define the point \(t^+_n\) by the relation \(S_n(t^+_n)=\sup_{t\in K} S_n(t)\). Such functional of a random field is of particular interest in mathematical statistics (the case of logarithmic likelihood function and \(K\) as a set of parameters). There are given rough asymptotics and exact local and integral theorems on large deviations of the maximum point \(t^+_n\) of a random field \(S_n\). For previous works on exponential bounds for maximum likelihood estimators see \textit{I. A. Ibragimov} and \textit{R. Z. Khas'minskij} [``Statistical estimation. Asymptotic theory''(1981; Zbl 0467.62026)] and \textit{M. Eh. Radavichyus} [Sov. Math., Dokl. 27, 127-131 (1983); translation from Dokl. Akad. Nauk SSSR 268, 551-556 (1983; Zbl 0528.62021)] and several earlier papers of the authors.