ep.bnpostmcmc: Posterior edge probabiities.

In rjbgoudie/structmcmc: Structural inference for Bayesian Networks and Variable Selection.

Description

Computes the edge probabilities implied by the MCMC samples contained in the 'bnpostmcmc' object x.

Usage

  ## S3 method for class 'bnpostmcmc'
 ep(x, nbin = 1, start, end,
    method = "et", verbose = F, ...)

Arguments

x	An object of class 'bnpostmcmc'
nbin	A numberic vector of length 1 specifying the number of bins into which to divide the MCMC samples. The edge probabilities are computed separately for each bin.
start	...
end	...
method	Either "et" (the default), "flatten" or "tabulate". If the edge totals are not available, method "tabulate" is used if possible. Only "flatten" is available if nbin != 1.
verbose	...
...	Further arguments passed to ep.parental.list() for method = "flatten", or ep.table() for method = "tabulate"

Value

if nbin == 1: A matrix of class 'ep' with entry (i,j) containing the probability of an edge from node i –> j if nbin > 1: A list of class ep.list, containing matrices as described above for each of the nbin bins into which the parental.list was split

See Also

ep, ep.bnpostmcmc.list

Examples

x1 <- factor(c(1, 1, 0, 1))
x2 <- factor(c(0, 1, 0, 1))
dat <- data.frame(x1 = x1, x2 = x2)

prior <- function(net) 1
initial <- bn(c(), c())

sampler <- BNSampler(dat, initial, prior)
samples <- draw(sampler, n = 50)
mpost <- bnpostmcmc(sampler, samples)

ep(mpost)

rjbgoudie/structmcmc documentation built on Nov. 3, 2020, 3:41 a.m.