mcmc_bite {bite}

MCMC algorithm

Description

Implements Markov chain Monte Carlo sampling for trait evolution models

Usage

mcmc_bite(model, log.file = "bite_mcmc.log", sampling.freq = 1000,
  print.freq = 1000, ncat = 1, beta.param = 0.3, ngen = 5e+06,
  burnin = 0)

Arguments

model

an object of class "jive" or other objects from the bite package (see details)

log.file

name of the output file that will store the log of MCMC chain

sampling.freq

sampling frequency of the MCMC chain (how often chain will be saved into output file

print.freq

printing frequency of the MCMC chain (how often chain will be printed in the R console). Setting it to 0 will suppress every printed message

ncat

number of classes for thermodynamic integration (see details)

beta.param

beta value to define classes for thermodynamic integration (see details)

ngen

number of generation in MCMC chain

burnin

a burning phase of MCMC chain (has to be specified for thermodynamic integration)

Details

This function runs MCMC sampling on jive object make_jive or objects describing other models of the bite package. The jive object contains both the dataset and set of model to be used in MCMC. This function implements both a conventional MCMC and an MCMC with thermodynamic integration. The latter option is turned off by default and can be changed by setting ncat to values > 1. The recommended ncat for TI is 10. When setting ncat > 1, make sure to specify burning. As a rule of thumb set burning to 1/10 fraction of ngen.

Value

Generates a log file in the users filespace at the path defined by log.file

Author(s)

Theo Gaboriau, Anna Kostikova, Daniele Silvestro, and Simon Joly

Examples

 ## Load test data
 data(Anolis_traits)
 data(Anolis_tree)
 data(Anolis_map)
 
 ## Run a simple MCMC chain
 set.seed(300)
 my.jive <- make_jive(phy = Anolis_tree, traits = Anolis_traits[,-3],
  model.priors = list(mean = "BM", logvar= c("OU", "root")))
 bite_ex <- tempdir()
 logfile <- sprintf("%s/my.jive_mcmc.log", bite_ex)
 mcmc_bite(model = my.jive, log.file=logfile,
 sampling.freq=10, print.freq=0, ngen=1000) 
  
 ## Run an MCMC chain with thermodynamic integration
 logfile <- sprintf("%s/my.jive_mcmc_TI.log", bite_ex)
 mcmc_bite(my.jive, log.file=logfile, ncat=10, 
  sampling.freq=10, print.freq=100, ngen=1000, burnin=10)