Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: st : Prediction using xtlogit


From   "Arne Risa Hole" <arnehole@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: st : Prediction using xtlogit
Date   Mon, 5 Mar 2007 18:10:57 +0000

While it is true that one cannot calculate the probabilities based on
a fixed effects logit model unless one makes the assumption that the
fixed effects are zero (since the fixed effects are not actually
estimated) this is not the case with the random effects logit model.
After all -gllapred- is able to calculate the probabilities without
imposing the assumption that the random effects are zero, so this does
not, I believe, explain why it is not possible with -xtlogit, re-

Arne

On 05/03/07, SamL <saml@demog.berkeley.edu> wrote:
Actually, I am not sure the fixed effects logit explicitly estimates what
one needs to have estimated to produce the predicted values you are
seeking.  I think you are estimating the random effects model, and I am
not sure this translates to that model.

But, the fixed effects logit sweeps the intercepts out, and estimates the
parameters for the variables that vary.  Lacking an intercept for a case,
I don't believe one can calculate a predicted value for the case.  It may
be that the random effects model does something similar, treating the
errors in the way that the fixed effects logit model treats the
person-specific intercepts.

But, I've been following this thread to see if someone mentioned this, and
I don't have access to my econometrics texts at the moment (while
traveling overseas), so I cannot do a quick check of my own resources to
confirm this.  But, it would explain the absence of a predicted value
"switch" on the post-estimation command for xtlogit, I suppose.

Hopefully others will weigh in to clarify.

HTH
Sam



On Mon, 5 Mar 2007, Arne Risa Hole wrote:

> Jeph,
>
> One way to speed up the estimation would be to run the model first
> using -xtlogit- and then use the estimated parameters as starting
> values for -gllamm-. Since you are estimating the same model using
> only slightly different approaches the coefficients ought to be very
> similar and therefore -gllamm- *should* only need a couple of
> iterations to converge.
>
> I agree that it doesn't sound like too much extra work is needed to
> calculate the probabilities using xtlogit.ado as a starting point but
> perhaps I'm ignoring something here..
>
> Arne
>
> On 05/03/07, Jeph Herrin <junk@spandrel.net> wrote:
> > Arne,
> >
> > Thanks. I considered this, but -gllamm- takes much longer
> > to converge (using quadrature) and I need to loop over many
> > models; the difference in estimation time would be measured
> > in days.
> >
> > It is also curious that -xtlogit- can't provide these
> > probabilities directly; it must estimate the random effects,
> > right?
> >
> > thanks,
> > Jeph
> >
> >
> >
> > Arne Risa Hole wrote:
> > > Hi Samuele and Jeph,
> > >
> > > If you estimate the random effects logit model using -gllamm- rather
> > > than -xtlogit- you can calculate the probabilities with non-zero
> > > random effects as well as the random effects themselves using
> > > -gllapred-. See ch. 9 in "A Handbook of Statistical Analyses using
> > > Stata" by Sophia Rabe-Hesketh and Brian Everitt for an example.
> > >
> > > Hope this helps.
> > >
> > > Arne
> > >
> > > On 05/03/07, Jeph Herrin <junk@spandrel.net> wrote:
> > >> I posted this same question *twice* this week, with no answer
> > >> yet. Please let me know if you learn something.
> > >>
> > >> thanks,
> > >> Jeph
> > >>
> > >>
> > >> Samuele Biasiolo wrote:
> > >> > After an xtlogit regression, using random effect, there is any
> > >> method to calculate probability of a positive outcome assuming random
> > >> effect is NOT zero?
> > >> >
> > >> > Thanks for help.
> > >> >
> > >> > Samuele
> > >> >
> > >> >
> > >> > ------------------------------------------------------
> > >> > Passa a Infostrada. ADSL e Telefono senza limiti e senza canone Telecom
> > >> > http://click.libero.it/infostrada4marz07
> > >> >
> > >> >
> > >> >
> > >> > *
> > >> > *   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/
> > >>
> > > *
> > > *   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/
> >
> *
> *   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/

*
*   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/



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