A suite of tools for denoising and analyzing one-dimensional signals with trend filtering. Trend filtering constructs a piecewise polynomial estimate (of any order) for the signal with knots chosen adaptively from the observed data. Hyperpameter(s) can be optimized by Stein's unbiased risk estimate or V-fold cross validation with a customizable loss function. Methods are also included for trend filtering uncertainty quantification and a generalized "relaxed trend filtering" estimator.
The trendfiltering package can be installed from GitHub using the command
below.
install.packages("remotes")
remotes::install_github("capolitsch/trendfiltering")-
Politsch et al. (2020a). Trend Filtering – I. A modern statistical tool for time-domain astronomy and Astronomical Spectroscopy. Monthly Notices of the Royal Astronomical Society, 492(3), p. 4005-4018. [Publisher] [arXiv] [BibTeX].
-
Politsch et al. (2020b). Trend Filtering – II. Denoising astronomical signals with varying degrees of smoothness. Monthly Notices of the Royal Astronomical Society, 492(3), p. 4019-4032. [Publisher] [arXiv] [BibTeX].