Get a slice of a mcmclist faster

Say you sampled a large amount of parameters from a posterior and stored the parameters in a mcmclist object. In order to analyze your results, you want to slice the mcmclist and only select a subset of the parameters. One way to do this is:


where mcmcsample is your mcmclist with the posterior sample and i the parameter you are interested in. Turns out, this function is very slow. A faster way is to use this little function:

getMCMCSample <- function(mcmclist,i){
  chainext <- function(x,i) return(x[,i])
  return(as.mcmc.list(lapply(mcmclist, chainext, i = i)))


# Run a toy model
counts <- c(18,17,15,20,10,20,25,13,12)
outcome <- gl(3,1,9)
treatment <- gl(3,3)
posterior1 <- MCMCpoisson(counts ~ outcome + treatment)
posterior2 <- MCMCpoisson(counts ~ outcome + treatment)
posterior3 <- MCMCpoisson(counts ~ outcome + treatment)
mcmclist <- mcmc.list(posterior1,posterior2,posterior3)

system.time(getMCMCSample(mcmclist, 2:3))

The build-in way takes 0.003 sec on my machine, while getMCMCSample gets it done in 0.001. For this little example, the difference is negligible. But as it turns out, for large posterior samples, it really makes a difference.