One possibility is that there is one sample in which the irls algorithm has a very hard time converging and thus takes a very long time.
hope this helps,
Maarten
-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands
visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434
+31 20 5986715
http://home.fsw.vu.nl/m.buis/
-----------------------------------------
-----Original Message-----
From: [email protected] [mailto:[email protected]]On Behalf Of Richard Sperling
Sent: donderdag 7 juni 2007 12:51
To: [email protected]
Subject: st: Monte Carlo run times with different seeds
I need help in understanding something I've observed.
I'm running a Monte Carlo simulation of the IRLS (iteratively
reweighted least squares estimator) on a nonlinear functional form.
If I use different seeds for generating the "true" data (seed = 123)
and the simulation (seed = 56203), then the runtime under the GUI
version of Stata MP 9.2 is about 24 minutes. If, however, I use the
same seeds to generate the true data and run the simulation (seed =
56203), then the runtime is 1 hour and 48 minutes. The run times are
slightly less when using the console version of Stata MP 9.2. I don't
understand why the seeds matter to such an extent.
I will be very appreciative if someone would enlighten me.
By the way, I ran the simulations on an 24" Apple iMac with an Intel
2.33GHz Core 2 Duo processor running Mac OS X 10.4.9.
Thanks,
Richard
*
* 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/
*
* 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/
