dclone (version 2.1-1)

bugs.parfit: Parallel computing with WinBUGS/OpenBUGS

Description

Does the same job as bugs.fit, but parallel chanis are run on parallel workers, thus computations can be faster (up to 1/n.chains) for long MCMC runs.

Usage

bugs.parfit(cl, data, params, model, inits=NULL, n.chains = 3, 
    seed, program=c("winbugs", "openbugs", "brugs"), ...)

Arguments

cl

A cluster object created by makeCluster, or an integer, see parDosa and evalParallelArgumen

data

A named list or environment containing the data. If an environment, data is coerced into a list.

params

Character vector of parameters to be sampled.

model

Character string (name of the model file), a function containing the model, or a or custommodel object (see Examples).

inits

Specification of initial values in the form of a list or a function, can be missing. If this is a function and using 'snow' type cluster as cl, the function must be self containing, i.e. not having references to R objects outside of the

n.chains

Number of chains to generate, must be higher than 1. Ideally, this is equal to the number of parallel workers in the cluster.

seed

Vector of random seeds, must have n.chains unique values. See Details.

program

The program to use, not case sensitive. See bugs.fit.

...

Other arguments passed to bugs.fit.

Value

• An mcmc.list object.

encoding

UTF-8

Details

Chains are run on parallel workers, and the results are combined in the end. The seed must be supplied, as it is the user's responsibility to make sure that pseudo random sequences do not seriously overlap. The WinBUGS implementation is quite unsafe from this regard, because the pseudo-random number generator used by WinBUGS generates a finite (albeit very long) sequence of distinct numbers, which would eventually be repeated if the sampler were run for a sufficiently long time. Thus it's usage must be discouraged. That is the reason for the warning that is issued when program = "winbugs". OpenBUGS (starting from version 3.2.2) implemented a system where internal state of the pseudo random number generator can be set to one of 14 predefined states (seed values in 1:14). Each predefined state is 10^12 draws apart to avoid overlap in pseudo random number sequences.

See Also

Sequential version: bugs.fit

Examples

## fitting with WinBUGS, bugs example
if (require(R2WinBUGS)) {
data(schools)
dat <- list(J = nrow(schools), 
    y = schools$estimate, 
    sigma.y = schools$sd)
bugs.model <- function(){
       for (j in 1:J){
         y[j] ~ dnorm (theta[j], tau.y[j])
         theta[j] ~ dnorm (mu.theta, tau.theta)
         tau.y[j] <- pow(sigma.y[j], -2)
       }
       mu.theta ~ dnorm (0.0, 1.0E-6)
       tau.theta <- pow(sigma.theta, -2)
       sigma.theta ~ dunif (0, 1000)
     }  
param <- c("mu.theta", "sigma.theta")
SEED <- floor(runif(3, 100000, 999999))
cl <- makePSOCKcluster(3)
if (.Platform$OS.type == "windows") {
sim <- bugs.parfit(cl, dat, param, bugs.model, seed=SEED)
summary(sim)    
}
dat2 <- dclone(dat, 2, multiply="J")
if (.Platform$OS.type == "windows") {
sim2 <- bugs.parfit(cl, dat2, param, bugs.model, 
    program="winbugs", n.iter=2000, n.thin=1, seed=SEED)
summary(sim2)
}
}
if (require(BRugs)) {
## fitting the model with OpenBUGS
## using the less preferred BRugs interface
sim3 <- bugs.parfit(cl, dat2, param, bugs.model, 
    program="brugs", n.iter=2000, n.thin=1, seed=1:3)
summary(sim3)
}
if (require(R2OpenBUGS)) {
## fitting the model with OpenBUGS
## using the preferred R2OpenBUGS interface
sim4 <- bugs.parfit(cl, dat2, param, bugs.model, 
    program="openbugs", n.iter=2000, n.thin=1, seed=1:3)
summary(sim4)
}
stopCluster(cl)
## multicore type forking
if (require(R2OpenBUGS) && .Platform$OS.type != "windows") {
sim7 <- bugs.parfit(3, dat2, param, bugs.model, 
    program="openbugs", n.iter=2000, n.thin=1, seed=1:3)
summary(sim7)
}