Bayesian resources

Petr Keil
January 2016

Software

OpenBUGS www.openbugs.net

  • Successor of (now discontinued) WinBUGS.
  • The software that made Bayesian analysis popular.
  • Has a GUI, but can also be commanded from R.
  • Great set of example models.
  • Useful for spatial analysis.
  • Difficult to debug, somewhat unpredictable.

R packages for OpenBUGS

  • BRugs - gives full control over OpenBUGS
  • R2OpenBUGS - simpler and more user-friendly

Software

JAGS mcmc-jags.sourceforge.net/

  • Just Another Gibbs Sampler.
  • Reliable, cross-platform.
  • Uses the BUGS language (almost identical to OpenBUGS).
  • Can be commanded from within R.

R packages for JAGS

  • rjags - gives full control over JAGS
  • R2jags - simpler and more user-friendly

Software

STAN - mc-stan.org

  • The newest and fastest one.
  • Can be commanded from R, Python or Linux shell.
  • Does not use the BUGS language, but the code is similar.
  • The future of MCMC.

R packages for STAN

  • RStan

Software

INLA - www.r-inla.org

  • Integrated Nested Laplace Approximation
  • Very fast
  • Whole different universe – not treated in this course.
  • Not as flexible as MCMC (yet?).

R packages for INLA

  • RINLA

Software

Other useful R packages - see the CTAN taskview

  • coda handles and summarizes MCMC output.
  • MCMCglmm simple Bayesian GLM fitting.
  • spBayes and geoRglm for spatially explicit modelling.
  • hSDM Hierarchical Species Distribution Modelling.
  • ggmcmc pretty output visualization in ggplot2 style.

Books