RE: st: RE: Error Message: r3900 - Problem with Memory of my Computer system ?!

["Pascal Stock" <pascalstock@freenet.de>]

Well, the if runiform()<.01 command causes a drop in observations. The
result is the uselessness of the regression results as the coefficients are
meaningless. Hence I looked for a manual adjustment to the
heteroscedasticity in the Heckman two-step probit selection model to obtain
efficient standard errors. A literature says that bootstrapping is a
solution to be used for the selection and structural equation. Hence I will
see if this works and if I can get STATA 11 from the Graduate School.

Kind regards and enjoy your Sunday

Pascal

____________________________________________
Pascal Stock

Quadrat N4, 8-9
68131 Mannheim
Germany

Phone: +49 (0) 621 - 16753 - 80
Fax: +49 (0) 621 - 16753 - 81
Mobile phone: +49 (0) 178 - 4981887
Email: pascalstock@freenet.de

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Martin Weiss
Sent: Sonntag, 16. Mai 2010 13:31
To: statalist@hsphsun2.harvard.edu
Subject: RE: st: RE: Error Message: r3900 - Problem with Memory of my
Computer system ?!


<>

To implement Kit`s proposal, Pascal could qualify his -cmp- call via

***********
if runiform()<.01
***********

Since he said he has 10.1, this should work.

HTH
Martin


-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Kit Baum
Sent: Sonntag, 16. Mai 2010 13:29
To: statalist@hsphsun2.harvard.edu
Subject: re:st: RE: Error Message: r3900 - Problem with Memory of my
Computer system ?!

<>
Pascal said

Martin gave me the extremely helpful advice to use the "cmp" command to
implement a three-stage selection model. So far it works fine. However, when
modeling the three-stage system with two selection equations and 1,000,000
observations I get the following error message when cmp tries to estimate
the combined model:

<istmt>:  3900  unable to allocate real <tmp>[1224521,46]
r(3900);

I already allocated the maximum memory (1000m) and matsize (10,000) to
STATA. I still get this error message.

So you have an idea how to solve the memory problem?

By my calculations that single matrix is about 4.2 Gb, so unless you have
access to a system where you can allocate
5 or 6 Gb to Stata, you have no hope of solving this problem. What happens
if you try to estimate the model with
10000 observations rather than a  million?

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


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/




*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/