Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: RE: xtmelogit


From   "JVerkuilen (Gmail)" <jvverkuilen@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: RE: xtmelogit
Date   Wed, 14 Nov 2012 09:45:32 -0500

On Wed, Nov 14, 2012 at 8:50 AM, Szabo S.M. <S.M.Szabo@soton.ac.uk> wrote:
> Thank you for getting back. This is a large merged household level dataset with over 650,000 observations over 38 countries. There two levels only - household level and contextual (country) level.
>
> Thank you for the valuable feedback. I think I will give Stata another go...

In general the execution time of a model is proportional to #params^2
+ #observations. With your huge dataset this means you'll have long
estimation times on any software you use.

Sounds like you actually have three levels, not two, country and
household. It's not obvious that countries are comparable either, so
what you may want to do is try analyzing countries separately.

Regardless with that huge a number of responses, you would probably
benefit by randomly (or stratified randomly) subsampling down to a
much more manageable dataset for your specification search. This gives
you the benefit of working on a smaller dataset and also lets you do
real cross validation of your model.
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


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