问题
I have a function call (to jags.parallel
) that works when given a numerical argument like n.iter = 100
but fails when the argument uses a variable value, n.iter = n.iter
. This looks like it might be a bug in jags.parallel
A minimal reproducible example of the error:
library(R2jags)
model.file <- system.file(package="R2jags", "model", "schools.txt")
J <- 8.0
y <- c(28.4,7.9,-2.8,6.8,-0.6,0.6,18.0,12.2)
sd <- c(14.9,10.2,16.3,11.0,9.4,11.4,10.4,17.6)
jags.data <- list("y","sd","J")
jags.params <- c("mu","sigma","theta")
jags.inits <- function(){
list("mu"=rnorm(1),"sigma"=runif(1),"theta"=rnorm(J))
}
Then this works:
jagsfit.p <- jags.parallel(data=jags.data, inits=jags.inits, jags.params,
n.iter=5000, model.file=model.file)
But this does not:
n.iter=5000
jagsfit.p <- jags.parallel(data=jags.data, inits=jags.inits, jags.params,
n.iter=n.iter, model.file=model.file)
Giving the error:
Error in checkForRemoteErrors(lapply(cl, recvResult)) :
3 nodes produced errors; first error: object 'n.iter' not found
I gather this has something to do with not exporting the variable n.iter
to the cluster, but it is not clear what parallel engine jags.parallel is using. Is there any way to trick R to evaluate n.iter
before passing it to the function?
回答1:
do.call()
is a great go-to friend in situations like this because (from ?do.call
):
If 'quote' is 'FALSE', the default, then the arguments are evaluated (in the calling environment, not in 'envir').
I confirmed that the following works, producing results that match your jagsfit.p
through all digits displayed by the result object's print method:
jagsfit.p2 <- do.call(jags.parallel,
list(data=jags.data, inits=jags.inits, jags.params,
n.iter=n.iter, model.file=model.file))
来源:https://stackoverflow.com/questions/16723036/strange-jags-parallel-error-avoiding-lazy-evaluation-in-function-call