Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Re: ram memory: ms fix does not work (surprised?)


From   Roberto Mura <roberto.mura@mbs.ac.uk>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Re: ram memory: ms fix does not work (surprised?)
Date   Tue, 07 Nov 2006 14:42:15 +0000

Dear Kit. Thank you very much for your kind reply. I appreciate that.
Indeed I am quite familiar with your papers.
Please allow me to briefly clarify the following: I have an
unbalanced panel with 10361 obs, 1039 firms and T is 17 (UK listed
non financial firms). That error message previously reported referred
to a lag specification t-2 t-5; however, even with a more
parsimonious specification with lags t-2 - t-3 as instruments the
error message still crops up:
xtabond2_mata():  3900  unable to allocate real <tmp>[35360,749]
                  <istmt>:     -  function returned error

Also, please let me clarify that I am running a GMM-SYS not a DIFF.
The xtabond2 however is just one of the manifestations of the
"contiguous memory" problem.
It becomes an issue again when, for instance, I try to work with all
Compustat firms.
Thank you again for your comments
Kind regards
R


At 13:22 07/11/2006, you wrote:
Others have suggested technical tweaks to your computer. I would
suggest that if you are trying to create 1049 GMM Instruments in
xtabond2, you probably want to rethink your strategy. If your T is
double-digit, letting xtabond2 (or any implementation of Arellano-
Bond) make all the possible instruments is a very bad idea. Use the
lags suboption to control how far back it goes to generate
instruments -- in my own work lags 3-5 are usually enough. Allowing
xtabond2 to use lags 3-17 is a recipe for disaster, even if you don't
run into problems with XP's memory mismanagement.

For comparison, an Apple xServe with 6 Gb RAM -- a "last-generation"
machine, not one of the new Intel Macs --  is running Stata as a 32-
bit app (due to the operating system) although the machine is 64-bit.
The next release of Mac OS X will allow 'true' 64-bit apps (as
Solaris or some Linuxes do). On that machine, I can get away with

set mem 2250m

with no problem.

Kit Baum, Boston College Economics
http://ideas.repec.org/e/pba1.html
An Introduction to Modern Econometrics Using Stata:
http://www.stata-press.com/books/imeus.html


On Nov 7, 2006, at 2:33 AM, Roberto wrote:

This is a huge issue as I cannot run the xtabond2 estimations I need
to, since the system runs out of memory:
   xtabond2_mata():  3900  unable to allocate real <tmp>[35360,1049]
                   <istmt>:     -  function returned error
*
*   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/
********************************************************************
Dr. Roberto Mura
Lecturer in Finance
The Manchester Accounting & Finance Group
Manchester Business School
University of Manchester
Crawford House, Oxford Road
Manchester
M13 9PL
UK
Room M37
tel: +44 (0) 161 275 0120
fax: +44 (0) 161 275 4023
home page: http://www.personal.mbs.ac.uk/roberto-mura/
e-mail: roberto.mura@mbs.ac.uk
            roberto.mura@manchester.ac.uk
********************************************************************

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index