/working/auteur.extended/man/tracer.Rd

http://github.com/eastman/auteur · Unknown · 39 lines · 35 code · 4 blank · 0 comment · 0 complexity · 41c1df67e53d8dc8133dd0538a8d09d4 MD5 · raw file

  1. \name{tracer}
  2. \alias{tracer}
  3. \title{plotting of Bayesian posterior samples }
  4. \description{
  5. Generates diagnostics for reversible-jump Markov sampling densities}
  6. \usage{tracer(base.log, lambdaK = log(2), Bayes.factor = NULL, burnin = 0.25, col.line = 1, pdf = TRUE, factor="M",...)}
  7. \arguments{
  8. \item{base.log}{a path to the stored summary logfile of the Markov sample}
  9. \item{lambdaK}{the shape parameter for the Poisson distribution used as a prior on the number of distinct rates in the tree (see \code{\link[stats]{rpois}})}
  10. \item{Bayes.factor}{a vector of two numbers, specifying a Bayes factor comparison between \code{j}- and \code{k}-rate models; alternatively, the argument
  11. can be \code{Bayes.factor=c(1,"multi")} in which a comparison between a global-rate and multiple-rate model is made}
  12. \item{burnin}{proportion of the chain to be treated as burnin (e.g., from 0 to 1); burnin is used for all plots but the trace of the log-likelihoods}
  13. \item{col.line}{a color to be used for density plots (both for a \code{Bayes.factor} comparison and for the density of mean rates across the Markov sample)}
  14. \item{pdf}{a logical switch that determines whether plots are written to .pdf (and stored within the stored Bayesian output)}
  15. \item{factor}{either a character ("k" or "M") or \code{NULL}, which determines how likelihood trace axis is defined: in thousands (\code{k}) or millions (\code{M}) of generations}
  16. \item{\dots}{further arguments to be passed to \code{\link{plot}}}
  17. }
  18. \value{
  19. if \code{pdf=TRUE}, a .pdf will be generated that includes the trace of log-likelihoods, the density of sampled mean rates, the
  20. correspondence between prior and posterior weights for \code{k}-rate models, and a Bayes factor comparison }
  21. \author{JM Eastman}
  22. \examples{
  23. # generate tree
  24. n=24
  25. phy=prunelastsplit(birthdeath.tree(b=1,d=0,taxa.stop=n+1))
  26. # simulate data
  27. dat=rTraitCont(phy=phy, model="BM", sigma=sqrt(0.1))
  28. # run reversible-jump MCMC for a short chain
  29. r=paste(sample(letters,9,replace=TRUE),collapse="")
  30. rjmcmc.bm(phy=phy, dat=dat, ngen=10000, sample.freq=10, prob.mergesplit=0.1, fileBase=r)
  31. # plot Markov sampled rates and other traces
  32. tracer(base.log=paste(paste("BM",r,"parameters",sep="."),paste("BM",r,"rjmcmc.log",sep="."),sep="/"), lambdaK=log(2), Bayes.factor=c(1,"multi"), burnin=0.25, pdf=FALSE, factor="k")
  33. ## PASTE UNCOMMENTED FOLLOWING LINE TO DROP DIRECTORIES CREATED BY RJMCMC
  34. # unlink(dir(pattern=paste(r)),recursive=TRUE)
  35. }