The Vervaat process in \(L_p\) spaces (Q5943596): Difference between revisions

Let \(E_n\) be the empirical distribution function based on a sample of size \(n\) from the uniform \((0,1)\) distribution, and let \(E_n^{-1}\) be the left-continuous inverse of \(E_n\). Moreover, set \(\beta_n^U=E_n-I\) and \(\gamma_n^U=E_n^{-1}-I\), where \(I\) denotes the identity function. Then the Bahadur-Kiefer process is defined by \(R_n^U=\beta_n^U+\gamma_n^U\), and the integrated Bahadur-Kiefer process \(V_n^U(t)=\int_0^t R_n^U(s) ds\), \(0\leq t\leq 1,\) is known as the [(0,1)-uniform] Vervaat process. After reviewing a number of results for \(V_n\) the following asymptotic almost sure result for the process \(Q_n^U=V_n^U-{1\over 2}\{\beta_n^U\}^2\) is proven: For any \(1\leq p< \infty\) \[ \lim_{n\to\infty}n^{1/2}{\|Q_n^U\|_p\over \|\beta_n^U\|_q^{3/2}}= {1\over\sqrt{3}}c_0(p)\quad \text{a.s.} \] where \(q=3p/2\) and \(c_0(p)=\sqrt{2}\{\Gamma((p+1)/2)/\sqrt{\pi}\}^{1/p}\).