R Packages


The R packages here perform a variety of analyses. I plan to upload them to CRAN once the relevant papers are published.

Unless otherwise mentioned, I am the maintainer for these packages. Email me if you identify any bugs or want to suggest features for future versions. The packages are all distributed under a GPL 3.0 version. This means you are free to distribute copies or modify the software. However any copies you distribute (modified or otherwise) must also be distributed under a GPL 3.0 or later.

PoissonPCA

Tianshu Huang & Toby Kenney

Dowload Version 1.0.1

PoissonPCA is an R package for fitting a corrected PCA to the (possibly transformed) latent Poisson means of a distribution.

 

Summary

Given a data matrix X where Xij∼ Po(Λij) are conditionally independent given Λ, this package estimates the covariance matrix of a transformation f(Λ), and from this estimates the principal components.

Documentation

The link below provides brief documentation of the functions provided by the package.

Documentation

SuRF

Lihui Liu

Dowload Version 1.0.2

SuRF is a variable selection method based on Forward Selection with Ranking by Subsampling.

 

Summary

Given a predictor matrix X and a response variable Y, this package aims to perform variable selection for a predictive generalised linear model. It does this in two stages: first a subsampling method with LASSO for ranking the variables, then a forward selection algorithm.

Documentation

The link below provides brief documentation of the functions provided by the package.

Documentation

Old Versions

  • Version 1.0.1
  • Version 1.0.0

  • AdequateBootstrap

    Toby Kenney

    Dowload Version 1.0.0

    AdequateBootstrap is an R package for performing the adequate bootstrap method, which reduces the bootstrap size based on model adequacy. Full details are in this paper.

     

    Summary

    Given a parametric model and some data, the adequate boostrap first calculates the bootstrap size at which a bootstrap sample has a 50% chance of rejecting the model adequacy test. It then uses bootstraps of this size to obtain confidence intervals for the parameters. The idea is that the confidence intervals should include the issue of model uncertainty.

    Documentation

    The link below provides brief documentation of the functions provided by the package.

    Documentation

    Note

    A c++ version of the program is available here.