Rolling windows
The rolling window approach partitions the univariate time series of
interest into a window of data points within which each indicator is
estimated. The window then ‘rolls’ along the time series one data point
at a time to update the indicator estimate and generate a new time
series of EWSs (Figure 2a). From this EWS time series, the Kendall’s Tau
correlation of the EWS against time is used to generate ‘warnings’
(Figure 2b). Specifically, if a strong Tau correlation is found, this
indicates an oncoming transition. The uniEWS function allows
the user to specify the window size as a percentage of the time series’
length and returns both the time series of EWSs and the estimated
Kendall’s Tau to be interpreted.