No label defined (Q4036738)

[For the entire collection see Zbl 0755.00022.] Let \(U_ 1\), \(U_ 2\) be a sequence of independent random variables with common uniform distribution on (0,1). Denote by \(F_ n(t)\) the right- continuous empirical distribution function based on \(U_ 1,\dots,U_ n\) and by \(G_ n(s)\) the empirical quantile function pertaining to \(F_ n\). Let \(\alpha_ n(s)=n^{1/2}(F_ n(s)-s)\) denote the empirical process and let \(\beta_ n(s)=n^{1/2}(G_ n(s)-s)\) denote the empirical quantile process. Define the Bahadur-Kiefer process \(R_ n(s)\) as \[ R_ n(s)=\alpha_ n(s)+\beta_ n(s)\quad (0\leq s\leq 1). \] The author obtains the following strong approximation of \(R_ n(s)\): there exists a sequence \(\{W_ i(t):-\infty<t<\infty\}\) \((i\geq 0)\) of independent standard Wiener processes such that \[ \left| R_ n(s)- n^{-1/2}\sum^ n_{i=1}W_ i\left({(s(1-s))^{1/2}\over n} W_ 0(n)\right)\right|=O(n^{-3/8}(\log n)^{3/4}(\log\log n)^{1/8})\;(0<s<1). \] Using this result, the author extends \textit{J. Kiefer}'s results [Ann. Math. Stat. 38, 1323-1342 (1967; Zbl 0158.370)] via simple arguments.