Model assessment for time series dynamics using copula spectral densities: a graphical tool (Q2001092)

The classical spectral analysis of time series is based on the autocovariance function (see, for example, the book by \textit{M. B. Priestley}, [Spectral analysis and time series. Volume 1: Univariate series. Volume 2: Multivariate series, prediction and control, Probability and Mathematical Statistics. London etc.: Academic Press (1981; Zbl 0537.62075)]). The autocovariance function does provide a complete description of the dependence of Gaussian processes, but it can completely miss dependencies in a non-Gaussian setting. This limitation attracted much attention, and new frequency domain tools that can capture non-linear dynamics are proposed. See, for example, the papers [\textit{H. Dette} et al., Bernoulli 21, No. 2, 781--831 (2015; Zbl 1337.62286)], [the second author et al., Bernoulli 22, No. 3, 1770--1807 (2016; Zbl 1369.62245)], [the first author et al., J. R. Stat. Soc., Ser. B, Stat. Methodol. 79, No. 5, 1619--1643 (2017; Zbl 06840449)] for more results and references. In the present paper, the authors utilize a method based on copula spectral densities to develop a graphical tool to determine suitable parametric models for time series. They provide a graphical tool that can indicate whether a chosen model accurately reflects the dependence present in the observed data. They provide a theoretical justification of the proposed method and show with the help of simulations that it can successfully distinguish between subtle differences in time series dynamics, including non-linear dynamics which result from GARCH and EGARCH models. The utility of the proposed tools is demonstrated by an application to modelling returns of the S\&P 500 stock market index.