Weak convergence of serial rank statistics under dependence with applications in time series and Markov processes (Q918102)

A class of linear serial rank statistics for the problem of testing white noise against alternatives of ARMA serial dependence was introduced by \textit{M. Hallin}, \textit{J.-F. Ingenbleek} and \textit{M. L. Puri} [Ann. Stat. 13, 1156-1181 (1985; Zbl 0584.62064)]. In that paper, the authors established the asymptotic normality of these statistics for ARMA processes contiguous to white noise. In the present paper, this asymptotic behavior is proved for \(\phi\)-mixing and strong mixing sequences of random variables. The authors show that contiguity is not necessary for the derivation of the asymptotic distribution theory. Their results also lead to applications in some Markov processes which are either geometrically ergodic or Doeblin recurrent, and to some ARMA processes.