Set up the rjMCMC sampler

Set up a list object to hold the data from the MCMC sampler, and generate starting values for all model parameters. This function is called internally by run_rjMCMC.

setup_rjMCMC(
  rj.input,
  n.burn,
  n.iter,
  n.chains = 3,
  p.split,
  p.merge,
  do.update = FALSE,
  start.values = NULL
)

Arguments

rj.input	Input dataset. Must be an object of class rjconfig.
n.burn	Number of MCMC iterations to treat as burn-in.
n.iter	Number of posterior samples.
n.chains	Number of MCMC chains. Defaults to 3.
p.split	Probability of performing a group split when model.select = TRUE.
p.merge	Probability of performing a group merge when model.select = TRUE.
do.update	Logical. Whether to update an existing sampler or set up a new one.
start.values	Starting values to use when updating an existing sampler and do.update = TRUE.

Details

In addition to split/merge moves, three other types of MCMC samplers are implemented in espresso to facilitate convergence and avoid getting stuck in local maxima: two data-driven samplers (type I and type II), in which proposals are informed by the “cues” present in the original data, and one independent sampler, in which proposals are drawn at random from a Uniform distribution bounded by range.dB (see read_data or simulate_data), with no dependence on current values.

See also

run_rjMCMC

Author

Phil J. Bouchet