On Friday, May 2, 2003, at 02:33 AM, Eswar wrote:
I have a dataset with about 5000 individuals, each with about 5
observations. I need to run panel data estimators like XTREGAR, XTGLS
and
calculated bootstrapped CIs on smaller regression samples. The data
set is
about 100MB. As I mentioned earlier I had a Pentium 4 256 MB computer
(with
virtual memory left to be determined by the system). A do file with
univariate and multivariate analyses about 20 lines takes a good 20 -30
minutes to run. Add to this the time needed to construct the .do file
(even
if is perform the debugging in a small subsample). Now after 1 run if I
realize that I want to make a change, it takes a whole length of time.
In
addition, I am not able to do anything useful while STAT is running.
I found from SUN representative that their Sunblade 150 can handle 5-6
gig
RAM easily as it is a 64 bit system. However, it costs a lot ( at
least 5000
dollars with 2 gig RAM). So here am I........... Hope this
information is
helpful to others.
Short of going to a 64-bit system, I see no reason why this sort of
analysis should be
cumbersome in a well-configured 32-bit system. 256 Mb of RAM is not
enough, especially with Windows' overhead, but if you take an Intel
machine, get rid of Windows (or partition the hard disk to make it a
dual-boot system) and install Red Hat Linux (or one of its competitors)
you will get _much_ better performance from the hardware, and Linux
will do a much better job of handling virtual memory than will Windows,
whatever the physical memory.
I routinely analyse panel datasets of ~400 Mb on a Sun Blade 100 which
cost under $1000 with 128 Mb RAM: I added a Gb of standard PC133
memory, and it deals with any XT analysis just fine (it could be bumped
to 2 Gb RAM, max). For that matter, so does my OS X Macintosh with 512
Mb RAM. But you can 'do it on the cheap' with an Intel box, for the
cost of some more RAM, a Linux Stata license, and a copy of Red Hat,
SuSE, Debian etc. And you'll no longer have to worry about 99.9% of
computer viruses...
Kit
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
