Reduction principles for quantile and Bahadur-Kiefer processes of long-range dependent linear sequences (Q946481)

The authors investigate quantile and Bahadur-Kiefer (BK) processes for long-range dependent (LRD) linear sequences \(X_i=\sum_{k=0}^\infty c_k \varepsilon_{i-k}\), where \((c_k)\) is regularly varying of index \(-1<\beta<-1/2\) and \((\varepsilon_k)\) is a centered i.i.d. sequence. For general BK-processes the usual approach for i.i.d. random variables \((X_i)\) is to approximate by uniform quantiles and to use the results valid for these. This does not work in the LRD case and the authors consider appropriate weight functions for the processes instead. Remarkably, unlike in previous studies, this allows to consider the uniform quantile and BK-processes on the whole interval \((0,1)\), although the tails of the quantile process can have a very erratic behavior. Various strong approximations are given, where the rate of convergence depends on whether \(\beta\leq-3/4\) or \(\beta>-3/4\) for the uniform quantile process, respectively \(\beta\leq-2/3\) or \(\beta>-2/3\) for the BK-processes. The study yields to some phenomena that cannot be obtained in the case of i.i.d. random variables \((X_i)\). These phenomena include weak convergence of the BK-process and a different almost sure behavior of the general and the uniform BK-processes.