Fast Rcpp/C++ translations of popular statistical hypothesis tests in R
Undoubtedly, R is a great analysis platform, but memory handling and speed is of great conern, especially in a 'big' data setting such as transcriptomics, proteomics, or public health data - just to name few domains of my interest. While the speed limitations may be partially addressed by parallelization, the R interface to C++, Rcpp provides the most elegant solution. The fastTest package bundles a growing set of statistical hypothesis testing tools, which are, except for ANOVA and correlation tests, almost 1:1 C++ translation of genuine R code and offer approximately 3- to 10-fold faster computation with minimal memory burden. The secondary motivation was to generate statistical hypothesis testing tools which return effect size statistics as well, at the moment the testing and effect size tools live in base R, stats and various packages with quite often incompatible interfaces. 'Standard' R functions and their fastTest counterparts are shown in the scheme below:
You may fetch the package with devtools:
devtools::install_github('PiotrTymoszuk/fastTest')
Credits to creators, contributors and maintainers of Rcpp, tidyverse, and furrr.
The package is available under a GPL-3 license.
The package maintainer is Piotr Tymoszuk.