NonParRolCor: An R package for estimating rolling correlation for two regular time series

Abstract

The R package NonParRolCor estimates rolling window correlations between two regular time series. The statistical significance estimated for the rolling correlation coefficients addresses effects due to multiple testing. This is done via Monte Carlo simulations by permuting one of the variables and keeping the other fixed. NonParRolCor uses parallel computing to improve computation time when statistical significance is estimated. NonParRolCor contains four functions for estimating and plotting the correlation coefficients for a single window-length or a band of window-lengths. NonParRolCor's functions are highly flexible, since they contain several parameters for controlling the estimation of correlation and the plot output. Some applications are presented to illustrate its use. © 2023 The Author(s)