Statalist The Stata Listserver


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

Re: st: Prediction using xtlogit


From   rgutierrez@stata.com (Roberto G. Gutierrez, StataCorp)
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Prediction using xtlogit
Date   Mon, 05 Mar 2007 12:34:14 -0600

There has been a lot of discussion as to how to obtain estimates of the random
effects in an -xtlogit, re- model.  Most recently, Arne Risa Hole
<arnehole@gmail.com> states:

> 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-

Predicting random effects is currently not possibly with -xtlogit, re-, and 
is this due to history and oversight.

History: -xtlogit, re- originally used standard Gaussian quadrature, an
estimation method not requiring direct estimates of the random effects in
order to perform the integration necessary to calculate the likelihood.
The current version now uses (by default) adaptive Gaussian quadrature, 
a method that does require estimating the random effects.

Oversight: No matter the estimation method, there is no reason that random
effects could not be "predicted" post-estimation, a la BLUP estimation after
-xtmixed-.

Resolution: This is a capability we'll certainly look into adding.

--Bobby
rgutierrez@stata.com
*
*   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