The R package pcpr implements Principal Component Pursuit (PCP), a robust dimensionality reduction technique, for pattern recognition tailored to environmental health data. The statistical methodology and computational details are provided in Gibson et al. (2022).

Installation

You can install the latest official CRAN release of pcpr with:

install.packages("pcpr")

The development version of pcpr can be installed from GitHub with:

# install.packages("pak")
pak::pak("Columbia-PRIME/pcpr")

pcpr can then be loaded and attached in your current R session as usual with

library(pcpr)

Getting help

Extensive documentation is available on our pkgdown website and offline within R. You can see the pcpr reference manual in R with:

help("pcpr")

A number of vignettes are available from within R. They can be browsed using:

browseVignettes("pcpr")

We recommend reading the vignettes in the following order:

1. Theory crash course, or if directly in R: vignette("theory-crash-course")
2. Quickstart, or if directly in R: vignette("pcp-quickstart")
3. Air pollution source apportionment with PCP, or if directly in R: vignette("pcp-applied")

Have a bug to report or question to ask? Open an issue on our GitHub.

Modeling overview

PCP algorithms model an observed exposure matrix