Markov chain Monte Carlo diagnostic (dMCMC). A set of functions that create multipanel plots to quickly evaluate MCMC output, and easily convert MCMC chain lists (returned by e.g. rjags) into handy-dandy latex tables. The motor behind these functions are the excellent R packages xtable and coda. Though, the main goal of dMCMC is to aggregate the most useful information in a single attractive graph or rapidly format a table for you. Which, if you work with numerous different Bayesian models daily, should prove usefull. One innovation that dMCMC adds to the mix is prior v.s. posteriors plots, which can be a crucial consideration, if one wan't to scrutenize choice of priors amongst other things.


Currently there isn't a release on CRAN, though there may one day be one. You can still download the zip or tar ball. Then decompress and run R CMD INSTALL on it, or use the devtools package to install the development version.

## Make sure your current packages are up to date
## devtools is required
install_github("dMCMC", "MarcoDVisser")


     cat( "
     model {
             for (i in 1:N) {
                     x[i] ~ dnorm(mu, tau)
             mu ~ dnorm(0, .0001)
             tau <- pow(sigma, -2)
             sigma ~ dunif(0, 100)
         } ",fill=TRUE)

     jags <- jags.model('examp.txt',
                       data = list('x' = rnorm(100,2,2),
                       n.chains = 4,
                       n.adapt = 100)

     mysamples <- coda.samples(jags, c('mu', 'tau'),100)


dMCMCs has the ability to find the priors directly in your BUGS/JAGS model file and translate these to their R equivalents. Here for the example above:


Which returns the accociated density function and translated parameters:

[1] "BUGS prior identified as:  mu~dnorm(0,.0001)"
[1] "BUGS prior translated to: dnorm"
[1] "dnorm"

[1] 0 1

more to follow

MarcoDVisser/dMCMC documentation built on May 7, 2019, 2:49 p.m.