A Bahadur-type representation for empirical quantiles of a large class of stationary, possibly infinite-variance, linear processes (Q749105)

X(n) is defined as \(\sum^{\infty}_{i=0}D(i)e(n-i)\), where the quantities e(n) are i.i.d. with \(E(| e(n)|^ A)\) finite for some positive A. For \(i\geq 1\), \(| D(i)| \leq ci^{-q}\) for some positive c, with \(q>1+2/A\). Let \(F_ n\) denote the empirical cumulative distribution function based on X(1),...,X(n), and let F denote the stationary distribution of X. For p in (0,1), \(X_{p,n}\) denotes the p th sample quantile of the sample X(1),...,X(n), and Q(p) is defined by the equation \(F(Q(p))=p\). We assume that the density f of the stationary distribution of X is bounded away from 0 and \(\infty\) in a neighborhood of Q(p). Then it is shown that \[ X_{p,n}=Q(p)+[p-F_ n(Q(p))]/f(Q(p))+R_ n\quad a.s., \] \[ where\quad R_ n=O(n^{- 3/4+g})\text{ for all } g>[A^ 2(8q-5)+2A(10q-9)-13]/[4(2Aq-A-1)^ 2]. \]