I'd said that I would send some example code. Here's code to do a
permutation test on the fake.bc data with a cluster of 8 nodes. We
first use scanone with the n.cluster argument, and then I spell out
the gory detail, which you might use to parallelize other things. (We
will be incorporating this sort of thing in upcoming versions of R/
qtl.)
data(fake.bc)
fake.bc <- calc.genoprob(fake.bc, step=1)
# requires the package rlecuyer
operm <- scanone(fake.bc, method="hk", n.perm=1000, n.cluster=8)
##############################
# doing it "by hand"
##############################
library(snow)
# set up 8 node cluster
cl <- makeCluster(8)
# set up random number generator
clusterSetupRNG(cl)
# each cluster needs to load the R/qtl package
clusterEvalQ(cl, require(qtl, quietly=TRUE))
# do the permutations
operm <- clusterCall(cl, scanone, fake.bc, method="hk", n.perm=125,
n.cluster=1)
# stop the clusters
stopCluster(cl)
# combine the results
for(i in 2:length(operm))
operm[[1]] <- c(operm[[1]], operm[[i]])
# keep just the combined results
operm <- operm[[1]]
karl