Multivariate Locally Stationary Wavelet Process Analysis with the mvLSW R Package

TL;DR

The paper introduces the R package mvLSW, which provides tools for analyzing multivariate locally stationary wavelet time series, including simulation, estimation, and inference methods, demonstrated through simulations and real data.

Contribution

The paper presents a comprehensive R package for multivariate LSW analysis, including simulation, estimation, and inference tools, with practical demonstrations.

Findings

Successful simulation of multivariate LSW time series.

Effective estimation of time-dependent EWS and coherence.

Application to real-world financial data.

Abstract

This paper describes the R package mvLSW. The package contains a suite of tools for the analysis of multivariate locally stationary wavelet (LSW) time series. Key elements include: (i) the simulation of multivariate LSW time series for a given multivariate evolutionary wavelet spectrum (EWS); (ii) estimation of the time-dependent multivariate EWS for a given time series; (iii) estimation of the time-dependent coherence and partial coherence between time series channels; and, (iv) estimation of approximate confidence intervals for multivariate EWS estimates. A demonstration of the package is presented via both a simulated example and a case study with EuStockMarkets from the datasets package. This paper has been accepted by the Journal of Statistical Software. Presented code extracts demonstrating the mvLSW package is performed under version 1.2.1.