A functional law of the iterated logarithm for tail quantile processes (Q1200015): Difference between revisions

Let \(0<U_{1,n}<\dots<U_{n,n}<1\) be the order statistics of a sample of \([0,1]\) uniform, independent r.v.'s and \(\{k_ n\), \(n\geq 1\}\) be a sequence of positive integers such that \(1\leq k_ n\leq n\), \(k_ n\uparrow\infty\) and \(k_ n/\log\log n\to 0\) as \(n\to\infty\). Consider the process \[ \zeta_ n(u)=(\log\log n)^{-1} k_ n(-\log U_{[k_ n,u],n}+\log(n^{-1} k_ n)). \] The main theorem claims that the sequence of functions \(\{\zeta_ n(u)\), \(0<u\leq 1\}\) is almost surely relatively compact in a given function space and the limit points are exactly characterized. By this theorem the tail behaviour of a quantile process can be characterized.