Quantile spectral processes: asymptotic analysis and inference (Q282565)

The analysis of autocorrelations and spectral densities as their frequency domain counterpart has proven to be a valuable tool for linear time series models. One major drawback is their inability of capturing nonlinear features of many time series models such as (G)ARCH and QAR models that have the same flat spectrum as strict white noise models. For this reason different dependency measures and the corresponding Fourier transformation (the spectrum) have recently been considered in a time series context.NEWLINENEWLINEFor example, \textit{H. Dette} et al. [Bernoulli 21, No. 2, 781--831 (2015; Zbl 1337.62286)] used the copula cross-covariances and the corresponding copula spectral density kernels to this purpose. The present paper derives rigorous asymptotic results for a corresponding class of smoothed rank-based cross-periodograms for a general class of possibly nonlinear time series opening the path to many statistical applications. The following applications are discussed theoretically and in a simulation study in this paper. First, asymptotic confidence intervals for the copula spectrum are derived. Secondly, asymptotic results for rank based spectral methods based on Spearman, Blomqvist or Gini autocovariances are provided.