fastrda: Fast Redundancy Analysis (RDA) with HPC C++ Backend

fastrda is a high-performance implementation of Redundancy Analysis (RDA) for R, written in C++ using Armadillo and OpenMP. It is designed for ecological, genomic, and other large multivariate datasets where computational efficiency is critical.

Features

High-performance C++ backend using Armadillo
OpenMP parallel computation
Standard and partial redundancy analysis (RDA)
Parallel permutation tests (anova() and anova_fastrda())
Prediction for new observations
Multiple score scaling options
Memory-efficient workspace modes
Publication-ready biplots
Familiar S3 interface (print(), summary(), predict(), scores(), plot(), biplot())

Installation

CRAN

install.packages("fastrda")

Development version

# install.packages("remotes")
remotes::install_github("zcebeci/fastrda")

Quick Example

library(fastrda)
library(vegan)

data(mite, mite.env)

# Hellinger transformation
Y <- decostand(mite, "hellinger")

# Environmental variables
X <- model.matrix(~ SubsDens + WatrCont, mite.env)[, -1]

fit <- fastrda(
  genotype = Y,
  environment = X,
  axes = 2,
  keep_workspace = "minimal"
)

summary(fit)

# Ordination plot
plot(fit)

Permutation test

anova(fit, permutations = 999)

Direct biplot

biplotrda(fit)

Workspace Modes

Mode	Recommended use
`"minimal"`	Permutation testing (recommended default)
`"compact"`	Prediction for new data
`"full"`	Prediction and permutation testing
`"none"`	Lowest memory usage

Documentation

The package vignette contains a complete workflow and additional examples.

vignette("fastrda")

Browse all available documentation:

help(package = "fastrda")

browseVignettes("fastrda")

Performance

Internal benchmarks on synthetic datasets containing up to 10,000 response variables showed median speedups of approximately 100× while maintaining numerical accuracy.

License

GPL (>= 3)