Locally Developed Software
The MRC Biostatistics Unit understands the importance of accessible software for putting new statistical methods into practice. Many BSU scientists write, or contribute to scientific software, which is typically made freely available and open-source.
This page lists software from BSU scientists that has been documented, tested and packaged in an accessible format. The Miscellaneous code section lists other downloadable code written by BSU scientists, typically to accompany a published paper.
BUGS is a language and various software packages for Bayesian inference Using Gibbs Sampling, conceived and initially developed at the BSU. Throughout its 20-year life span, BUGS has been highly influential in enabling the routine use of Bayesian methods in many areas. The current version, OpenBUGS, is open source, and offers improved sampling efficiency, modelling flexibility and interoperability with other software.
The newest development is MultiBUGS, a version of BUGS with an advanced parallel computation facility.
The ice package for Stata implements multiple imputation for missing data using chained equations, and includes major contributions by Ian White of the BSU and others. This work has led to improvements in methods for handling missing data in Stata itself. For more information, and instructions for downloading and installation, see Royston and White (Journal of Statistical Software, 2011).
Other Stata programs written by Ian White and other BSU authors is listed in the Stata Software page.
Bioinformatics and Statistical Genomics software
Bayesian Approaches to Clinical Trials and Health-Care Evaluation
See the Bayesian Approaches to Clinical Trials and Health-Care Evaluation page for code to accompany this popular book, that was partially written at the BSU.
Other R packages written by BSU members include:
- bcrm: Bayesian continual reassessment method (CRM) designs for Phase I dose-finding trials.
- PReMiuM: for Dirichlet process Bayesian clustering, or profile regression.
- flexsurv: flexible parametric models for time-to-event data, including the Royston-Parmar spline model, generalized gamma and generalized F.
- denstrip: for density strips and other plots for compactly illustrating distributions.
- ecoreg: for ecological regression using aggregate and individual data.
- R2GUESS: R interface to GUESS: GPU-enabled sparse Bayesian variable selection method for linear regression based analysis of possibly multivariate/correlated outcomes.
- GUESSFM: R package for fine mapping genetic associations in dense or imputed GWAS genotype data
- coloc: Colocalisation tests of two genetic traits
- GGMGSA: Multivariate Geneset Testing based on Graphical Models.
- mreg: Likelihood estimation for a longitudinal negative binomial regression model with missing outcomes.
- R2BGLiMS: An R interface to BGLiMS (Bayesian Generalised Linear Model Selection); a Java package for fitting survival and logistic Bayesian models under Reversible Jump model selection. The package can be downloaded from github by following the link, or installed directly in R after loading the ‘devtools’ package and entering ‘install_github’ (username=”pjnewcombe”, repo=”R2BGLiMS”). A Java JDK must also be installed.
- BASiCS: Bayesian Analysis of Single-Cell Sequencing data.
- ShrinkNet: Gene network reconstruction using global-local shrinkage priors.
- R2HESS: An R package that implements an efficient Bayesian approach for simultaneous association detection in OMICS datasets, with application to eQTL mapping in multiple tissues. Please click here to download the package: R2HESS_1.0.1.tar.gz
- Rwui: Rwui is a Java-based method to create a web interface for an R script. This allows scientific collaborators with no knowledge of the R statistical software to deploy novel statistical methods implemented in R. It has been used to create more than 1000 web applications over the last two years alone. For more information see the Rwui site, in particular the article by Newton and Wernisch (R News, August 2012).
- msm: multi-state modelling: The msm package for multi-state modelling in R is developed by Christopher Jackson. It implements general Markov and hidden Markov models for longitudinal data, especially data consisting of observations at arbitrary times. In its 15 year life span it has enabled the routine use of these methods in medical and other fields such as ecology, finance, social science and engineering. For more information see Jackson (Journal of Statistical Software, 2011).
- DeLorean: For the analysis of single cell transcriptomic data with uncertainty in the temporal dimension.
Other web-based applications written by BSU members include:
- AplusB: an investigative tool for the performance of A + B designs for phase I dose-escalation studies
- MetaAnalyser: a web application to visualise meta-analysis as physical weights on scales. See J. Bowden, C. Jackson, Weighing evidence ‘steampunk’ style via the Meta-Analyser, to be published in The American Statistician (arXiV). Code and R package also available on GitHub.
Other Miscellaneous code
For “unofficial” program code written by BSU scientists, typically written to implement a single method from a published paper, see the Miscellaneous code page.
This page also includes an archive of code developed by scientists that have left the BSU, and which has not been maintained since the authors left.