mrfrj: Reversible Jump for MRFs over different interaction...

In Freguglia/mrf2dbayes: Bayesian Analysis of Markov Random Fields on 2d Lattices

Reversible Jump for MRFs over different interaction structures

Description

A Metropolis-Hastings algorithm that allows jumps over spaces of different dimensions. This algorithm allows spaces with restrictions (zero equality) to be visitted, providing a model selection framework along with the Bayesian solution for inference within models.

Usage

mrfrj(
  z,
  llapprox,
  nsamples = 1000,
  init_theta = "zero",
  init_included = "zero",
  sdprior = 1,
  sdkernel = 0.005,
  sdbirth = 0.05,
  kernel_probs = c(4, 1, 1, 1, 1),
  logpenalty = log(prod(dim(z))),
  verbose = interactive()
)

Arguments

z	The observed random field.
llapprox	The likelihood approximation to be used.
nsamples	Number of MCMC samples.
init_theta	Initial values of the MCMC algorithm. Set to "zero" to automatically create a vector equal to zero with appropriate length or "pl" to use the Maximum Pseudolikelihood estimator of z as the initial value.
sdprior	Sample Deviation of the Normal distributions used as prior.
sdkernel	Sample Deviation of the Normal distributions of the transition kernel.

Value

An object of class mrfbayes_out

Freguglia/mrf2dbayes documentation built on Jan. 19, 2024, 11:21 p.m.