*****  To join INSNA, visit http://www.insna.org  *****

On 4/26/07, Tucker McGrimmon <[log in to unmask]> wrote:
> *****  To join INSNA, visit http://www.insna.org  *****
>
> Hi,
>
>   I am working with a large, but relatively sparse network, and trying
> to fit some ERGMs. There appears to be memory problems when using R. The
> typically error message R produces is
>
>   "Error: cannot allocate vector of size 222757 Kb"
>
> My understanding is that while the ERGM is running R runs out of
> virtually memory. Typically this could be solved by using a 64 bit
> machine and allowing it some time. But it also depends on R.
>
> I was wondering if anyone knew if R and the package Statnet were written
> with 32 bit or 64 bit hardware in mind. That is, is the memory problem
> above arising because I am running R & Statnet within windows on a 32
> bit computer or because R & Statnet were written as a "32 bit program"
> ie. software versus hardware problem or both?

You are just running out of memory. What else is running on the
machine at the same time? What does the code look like? Did you write
it yourself? The code might be doing things inefficiently (e.g.
allocating one array of size N and filling it, then doing passes over
it to perform various operations, instead of only stashing what's
actually minimally needed at the time). If the code really is
optimized to perfection, your dataset may be larger than you think it
is.

A program can't really be written "with 32 bit or 64 bit hardware in
mind". It can only be written to run on such hardware, or not. The
default these days is 32 bit, and so if you don't know what you
downloaded it's probably 32 bit. Further, if your machine is a 32 bit
machine then it definitely is a 32 bit binary--it would simply not
run.

There's not much of a solution here. Get a more souped up computer and
try again. Maybe building a cluster-computer would help you?

-Nick

_____________________________________________________________________
SOCNET is a service of INSNA, the professional association for social
network researchers (http://www.insna.org). To unsubscribe, send
an email message to [log in to unmask] containing the line
UNSUBSCRIBE SOCNET in the body of the message.