Local Gaussian cross-spectrum analysis

TL;DR

This paper extends local Gaussian spectral analysis from univariate to multivariate time series, enabling detection of non-Gaussian traits and local phenomena that traditional spectra might miss.

Contribution

It introduces a local Gaussian cross-spectrum for multivariate time series, enhancing the detection of non-Gaussian and nonlinear local structures.

Findings

Local Gaussian cross-spectrum coincides with the ordinary cross-spectrum for Gaussian series.

Peaks in the local Gaussian spectrum indicate potential nonlinear traits.

Method detects local phenomena undetectable by traditional spectral analysis.

Abstract

The ordinary spectrum is restricted in its applications, since it is based on the second order moments (auto and cross-covariances). Alternative approaches to spectrum analysis have been investigated based on other measures of dependence. One such approach was developed for univariate time series by the authors of this paper using the local Gaussian auto-spectrum based on the local Gaussian auto-correlations. This makes it possible to detect local structures in univariate time series that looks like white noise when investigated by the ordinary auto-spectrum. In this paper the local Gaussian approach is extended to a local Gaussian cross-spectrum for multivariate time series. The local Gaussian cross-spectrum has the desirable property that it coincides with the ordinary cross-spectrum for Gaussian time series, which implies that it can be used to detect non-Gaussian traits in the time series under investigation. In particular: If the ordinary spectrum is flat, then peaks and troughs of the local Gaussian spectrum can indicate nonlinear traits, which potentially might reveal local periodic phenomena that goes undetected in an ordinary spectral analysis.