wfinsinger/tapas: Trend And PeAkS analysis in paleo-ecological time series

Tapas proceeds through several steps to assess trend and peaks for paleoecological temporal time series. First, the time series can be virtually re-sampled to equal sampling intervals ("prebinning"). Second, the resampled time series is detrended using a smoothing function. Third, one 'global', or several 'local' 2-component Gaussian Mixture Model(s) (GMM) is/are fitted to the detrended data. One of the components is centred around 0, and the nth quantile of the gaussian distribution of that component is set to determine the threshold separating the noise from the signal. Eventually, figures can be produced to visualize the re-sampled data, the smoothed data, the detrended data, the GMMs, the thresholds, and the detected peaks (values beyond the noise threshold). To evaluate the suitability for peak detection of a temporal series, "tapas" embeds and uses the SNI.R script written by Ryan Kelly et al., and available in supplementary materials of Kelly et al., (2017) doi:10.1016/j.yqres.2010.07.011.

install.packages("remotes")
remotes::install_github("wfinsinger/tapas")