Convergence of integrals of uniform empirical and quantile processes (Q2366187)

Let \(\{U_ n, n \geq 1\}\) be a sequence of independent r.v.'s uniformly distributed on [0,1]. The uniform empirical process of \(U_ 1,U_ 2,\ldots,U_ n\) is defined by \[ e_ n(t)=n^{1/2} \bigl( E_ n(t)-t \bigr),\;0 \leq t \leq 1;\quad E_ n(t)=n^{-1}\sum_{1 \leq i \leq n} I[U_ i \leq t], \] and the uniform quantile process \(\tilde U_ n(t)\) by \[ \tilde U_ n(t) =n^{1/2} \bigl( t-U_ n(t) \bigr),\quad 0 \leq t \leq 1, \] where \(U_ n(t)=U_{k,n}\), \(k/(n+2)<t \leq (k+1)/(n+2)\), \(k=0,\ldots,n+1\), and \(U_{1,n} \leq U_{2,n} \leq \cdots \leq U_{n,n}\) denote the order statistics \(U_ 1,U_ 2,\ldots,U_ n\); \(U_{0,n}=0\), \(U_{n+1,n}=1\). Assume that \(q\) (a weight function) is a positive function on (0,1), i.e. \(\inf_{\delta \leq t \leq 1-\delta} q(t)>0\) for all \(0< \delta<1\). The authors give a description for \(L_ 1\)- functionals of \(e_ n/q\) and \(\tilde U_ n/q\) in the following theorems. Theorem 1.1. We assume that \(0<p<\infty\) and \(q\) is positive on (0,1). Then the following statements are equivalent: (i) We have \(\int^ 1_ 0 \bigl( t(1-t) \bigr)^{p/2}/q(t)dt<\infty\). (ii) There is a sequence of Brownian bridges \(\bigl\{ B_ n(t),\;0 \leq t \leq 1\bigr\}\) such that \(\int^ 1_ 0 | e_ n(t)-B_ n(t) |^ p/q(t) dt= o_ p(1)\). (iii) We have \(\int^ 1_ 0 | e_ n(t) |^ p/q(t)dt @>{\mathcal D}>>\int^ 1_ 0 | B_ n(t) |^ p/q(t)dt\), where \(\bigl\{ B_ n(t),\;0 \leq t \leq 1\bigr\}\) is a Brownian bridge. Theorem 1.2. We assume that \(0<p<\infty\) and \(q\) is positive on (0,1). Then the following statements are equivalent: (i) We have \(\int^ 1_ 0\bigl( t(1-t) \bigr)^{p/2}/q(t)dt<\infty\). (ii) There is a sequence of Brownian bridges \(\bigl\{B_ n(t),\;0\leq t\leq 1\bigr\}\) such that \(\int^ 1_ 0 | \tilde U_ n(t)-B_ n(t) |^ p/q(t)dt=o_ P(1)\). (iii) We have \(\int^ 1_ 0| \tilde U_ n(t) |^ p/q(t)dt @>{\mathcal D}>>\int^ 1_ 0 | B_ n(t) |^ p/q(t)dt\), where \(\bigl\{ B_ n(t),\;0 \leq t \leq 1 \bigr\}\) is a Brownian bridge.