[R] gls and memory allocation

Dave Depew ddepew at sciborg.uwaterloo.ca
Tue Aug 12 16:17:05 CEST 2008

Hi all,
A two part quick question regarding gls

1) I'd like to pass a formula to gnls from a gam(mgcv package). Is there 
a quick way to do this? I tried using

 >gnls(gam$call$formula, data=df, correlation=AC1) but it keeps 
outputting a message to the effect that I need x ~ y etc as the formula. 
I realize now that the $formula part of the gam is only sending x ~ s(y) 
and that this is probably the problem, but I cannot seem to find any 
part of the gam object that would pass the smooth terms as part of the 
formula. Any ideas or suggestions?

2) While running a gls with a rather large dataset (~ 13000 data points) 
it cannot seem to allocate ~ 685 MB for the vector. I've got a 2G RAM on 
windows XP (32 -bit I guess). I've cleared the memory as best I can, and 
specified the max allowable in the target line (i.e. --max-mem-size = 
2000MB) but still it does not want to run. Am I just running out of RAM? 
or should R be able to handle that size vector, and I just need to clean 
up the code?



More information about the R-help mailing list