Samuele Biasiolo wrote:
I'm trying to estimate a random effects logistic model. I used both
xtlogit and gllamm for comparison.
Running xtlogit and gllamm with the same number of integration points
give this results :
sigma_v is the sd of random effects.
xtlogit adaptive quadrature : log-likelihood=-8711.7129 sigma_v=0.6145
xtlogit standard quadrature : log-likelihood =-8695.7637 sigma_v=2.122
gllamm adaptive : log-likelihood=-8695.8048 sigma_v=2.1156
gllamm standard: log-likelihood=-8695.7633 sigma_v=2.122
I know that xtlogit and gllamm use differents method of adaptive
quadrature.
I'm surprise that standard quadrature perform better than adaptive,
especially in xtlogit.
--------------------------------------------------------------------------------
What makes you think that standard quadrature performs better than adaptive
quadrature?
Is it the increase in the log likelihood? I'm not so sure that that's the
best measure of this aspect of model fitting
Joseph Coveney
*
* 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/
