95
0

LDP application for Billingsley's example

Abstract

We consider a classical model discussed in Theorem 16.4 (Billingsley \cite{Bil}) concerning to an empirical distribution function F_n(t)=\frac{1}{n}\sum_{k=1}^nI_{\{\xi_k\le t\}}, of (ξk)i1(\xi_k)_{i\ge 1} - i.i.d. sequence of random variables, supported on the interval [0,1][0,1], with F(t)=P(ξ1t)F(t)=\mathsf{P}(\xi_1\le t) the continuous distribution function. We give a proof of Kolmogorov's exponential estimate \mathsf{P}\Big(\sup_{t\in[0,1]}|F_n(t)-F(t)|\ge \varepsilon\Big) \le 2\exp\Big\{-n[\varepsilon/8\log(1+\varepsilon^2/32)-\varepsilon/8 +(4/\varepsilon)\log(1+\varepsilon^2/32\Big)]\} with the help of which jointly with the large deviations technique, we establish a logarithmic asymptotic: for any T[F1(1/2),1)T\in[F^{-1}({1/2}),1) and any α(0,1/2)\alpha\in\big(0,{1/2}\big): \lim_{n\to\infty}\frac{1}{n^{1-2\alpha}} \log\mathsf{P}\bigg(\sup_{t\in[0,T]}n^\alpha\Big|F_n(t)-F(t)\Big|\ge \varepsilon\bigg)=-2\varepsilon^2.

View on arXiv
Comments on this paper