runjags (version 0.9.4-2)

combine.mcmc: Combine Two or More MCMC Objects With the Same Number of Chains Into One Longer MCMC Object

Description

Allows an MCMC object (with 1 or more chains) to be combined with another (or several other) MCMC object(s) representing extensions of the same simulation, to produce one MCMC object that contains the continuous combined Markov chains of the other MCMC objects. Also provides a safe way to thin a single MCMC object or list.

Usage

combine.mcmc(mcmc.objects=list(), thin=1, return.samples=NA, collapse.chains=FALSE)

Arguments

mcmc.objects

a list of MCMC objects, all with the same number of chains, or a single MCMC object or list. No default.

thin

an integer to use to thin the (final) MCMC object by, in addition to any thinning already applied to the objects before being passed to combine.mcmc. Ignored if return.samples is specified (!=NA). Default 1 (no additional thinning is performed).

return.samples

the number of samples to return after thinning. The chains will be thinned to as close to this minimum value as possible. Supersedes thin if both are specified. Ignored if niter(mcmc.objects) < return.samples. Default NA.

collapse.chains

option to combine all MCMC chains into a single MCMC chain with more iterations. Can be used for combining chains prior to calculating results in order to reduce the Monte Carlo error of estimates. Default FALSE.

Value

• A single MCMC object of length equal to the sum of the lengths of the input MCMC objects (if thin=1)

See Also

run.jags, testjags

Examples

# run a model, then extend the simulation and combine the two MCMC objects, with thinning to 5000 samples.

# Model in the JAGS format
model <- "model {
for(i in 1 : N){ #data# N
Y[i] ~ dnorm(true.y[i], precision); #data# Y
true.y[i] <- (m * X[i]) + c; #data# X
}
m ~ dunif(-1000,1000);
c ~ dunif(-1000,1000);
precision ~ dexp(1);
#monitor# m, c, precision
}"

# Simulate the data
X <- 1:100
Y <- rnorm(length(X), 2*X + 10, 1)
N <- length(X)

results1 <- run.jagsfile(model, n.chains=2, burnin=5000, sample=10000)
results2 <- run.jagsfile(model, inits=results1$end.state, burnin=0, sample=10000)

results <- combine.mcmc(list(results1$mcmc, results2$mcmc), return.samples=5000)

# Analyse the results
summary(results)