I have also noticed that SAS runs various binary type regression
models with complex error structures (though not standard logistic
regression) much faster than Stata for the medium-large data sets I
work with (approx 20,000 observations, 50 variables). Though I can't
recall the specific models now, when I get a chance, I document this
and summarize for the list.
Tim
On Wed, Jul 9, 2008 at 7:26 AM, Mitchell F. Berman <[email protected]> wrote:
> Thank you to Joseph for that ado file to test xtlogit, and to Dave for also
> checking the timing.
> This data set took 1.5 minutes on my computer, so the issue must be related
> to my data set.
>
> I don't understand the numerical methods well enough to know whether
> unbalancing in the panel variable affects the process. I have 2420 groups
> with between 1 and 2609 observations per group; avg 107. Dave's groups are
> more balanced.
>
> Standard logistic regression runs in seconds on my data-- though because the
> method is entirely different, this probably means little.
>
> I will check the link provided by Maarten-- perhaps there is some real
> difference between SAS and Stata concerning the methods. Stata has a
> de-identified version of my data set. I'm waiting to hear back.
>
> Mitchell Berman
> Columbia University
>
>
>
>
>
> *
> * 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/
