Re: st: FIRTHLOGIT with factor (categorical variables)

[Maarten buis <maartenbuis@yahoo.co.uk>]

--- On Tue, 21/9/10, Warren Chow wrote:
> But the following does not work:
> 
> firthlogit y i.x
> 
> Is there another version of the macro that supports factor
> variables, or an easy way to modify to code? I have tried
> the following:
> 
> xi : firthlogit y i.x
> But this is extremely slow.

That is the right solution, so if you want to use -firthlogit 
you'll just have to learn to live with the slowness. Moreover,
I am always left with a feeling of arbitrariness when it 
comes to the kind of penelized likehood used in -firthlogit-. 
I guess it won't matter too much in most practical application, 
but the logic doesn't sound too convincing to me (or I just do 
not understand it...). 

You could try -exlogistic- but i doubt whether that is going to
be quicker. Do not be fooled by the name though: exact methods, 
like -exlogistic-, sound great, but they are not as exact as the 
name suggest. On many properties that we care about, like 
coverage, the "exact" methods under perform compared to 
"non-exact" methods.

Alternatively, you can try to combine some categories of x1, so 
you can just use -logit-. As always, I prefer to simplify the 
model in cases like these, so that would be my prefered approach.

Hope this helps,
Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany

http://www.maartenbuis.nl
--------------------------


      

*
*   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/