Software – MRC Biostatistics Unit

Artistic image of computing cables

Overview

Our researchers routinely develop software to implement novel statistical methods, and make this software freely-available and open source.

We have a GitHub page which contains mirrors of the source code for packages that we currently maintain. This also includes code to reproduce analyses in published papers.

We have developed an array of Shiny applications, which can be browsed here in the index of MRC BSU Shiny Applications.

Open-source software

Many of our tools are freely available through GitHub, CRAN and other open-source platforms.

Our GitHub repositories include:

  • Statistical software packages
  • Research reproducibility code
  • Data analysis tools
  • Shiny applications
  • Methods developed through published research

Researchers can also explore a collection of BSU-developed Shiny applications and interactive tools.

The BUGS Project

BUGS is a language and various software packages for Bayesian inference Using Gibbs Sampling, conceived and initially developed at the BSU. Throughout its life span of over 30 years, BUGS has been highly influential in enabling the routine use of Bayesian methods in many areas.

Download the zip file and unzip this to the place where you want to install. The program will then be found in WinBUGS14/WinBUGS14.exe. This is the latest, unlocked version, and there is no need for an additional "patch" or "key". WinBUGS license text.

Download either the Windows installer or the zip file. The zip file should work if the the installer does not work - simply unzip it to the place where you want to install.  The program will then be found in OpenBUGS323/OpenBUGS.exe.  OpenBUGS is under the GNU General Public Licence version 3.

MultiBUGS is a more recent program, similar to OpenBUGS, but with facilities for parallel computation. This is hosted on GitHub. See the issues page for problem reports.

The BUGS Book: A Practical Introduction to Bayesian Analysis (Lunn, Jackson, Best, Thomas and Spiegelhalter, 2012) is a textbook about the BUGS language and software and its use for statistical modelling.

Two sample chapters are made available here:

Bugs book chapter 5 (pdf, 259.07 KB)
Bugs book chapter 9 (pdf, 421.8 KB)
  • A comprehensive set of code examples from the book (zip file). This includes the code in either ODC (the WinBUGS native file format) or HTML format.
  • A small selection of exercises (zip file).

While these programs are made available here, they are no longer actively developed.  For "help" questions about modelling in BUGS, please seek advice from the community via the public mailing list.

BUGS has inspired a large field of Bayesian modelling and "probabilistic programming" software.  Currently widely-used and actively-developed programs in this area include:

  • JAGS: (based on BUGS, but easily used across different platforms)
  • Stan: (with efficient gradient-based computation and a large community of users)
  • NIMBLE: (which allows the user to write and customise the sampling algorithms)
  • Turing: (based on the Julia language)

R packages

A large amount of the work at the BSU is done with the R statistical software. We have developed many R packages that are widely used.

  • msm: multi-state modelling of intermittently-observed data
  • coloc: Colocalisation tests of two genetic traits.
  • Mendelian Randomization: methods for performing Mendelian randomization analyses with summarized data.
  • BASiCS: Bayesian Analysis of Single-Cell Sequencing data.
  • GUESSFM: R package for fine mapping genetic associations in dense or imputed GWAS genotype data
  • BATSS: Defines operating characteristics of Bayesian Adaptive Trials considering a generalised linear model response via Monte Carlo simulations of Bayesian GLM fitted via integrated Laplace approximations (INLA).
  • MAMS: Designing multi-arm multi-stage studies with (asymptotically) normal endpoints and known variance.
  • flexsurv: Flexible parametric models for time-to-event data, including the Royston-Parmar spline model, generalized gamma and generalized F distributions.
  • DGP4LCF: Functionalities for analyzing high-dimensional and longitudinal biomarker data to facilitate precision medicine, using a joint model of Bayesian sparse factor analysis and dependent Gaussian processes

  • OnlineFDR: This package allows users to control the false discovery rate (FDR) or familywise error rate (FWER) for online multiple hypothesis testing, where hypotheses arrive in a stream.

  • vbvarsel - a variable bound, variable selection clustering algorithm.
MRC Biostatistics Unit and University of Cambridge logos