试图从大mcmc.list中提取内存已用尽

Question

I have problem trying to extract a single variable from big mcmc.list :

> x <- outRJ$mcmc[,'avail_int']
Error: cannot allocate vector of size 146.8 Mb

The outRJ$mcmc mcmc.list is big, 442 MB actually (contains data for 3 chains for like 960 variables, each 20000 iterations):

> object.size(outRJ$mcmc)/1024/1024
[1] 442.5508 #

But the single variable should have like 234 kB only! But the extraction of it seems to require much much more memory!

How can I extract from a big mcmc.list without requiring too much memory?

Answer 1

I think this should work:

 x <- as.mcmc.list(lapply(outRJ$mcmc, function(x) return(x[,'avail_int',drop=FALSE])))

Matt

Answer 2

The current subsetting code (in the function [.mcmc.list) is making a copy of each element of the list before subsetting it. I have modified the code in the development version of coda. In the meantime, Matt's workaround should do because it also avoids the redundant copy.