Pierre Azoulay wrote:
I am no expert on gllam, but I ran side by side what you propose and
xtpoisson accident op_75_79 co_65_69 co_70_74 co_75_79, ///
fe i(ship) offset(ln_service) irr
I get coefficient estimates that are close, but not the same.
If I type
xtpoisson accident op_75_79 co_65_69 co_70_74 co_75_79, ///
re i(ship) offset(ln_service) irr
Then I get close to the results given by gllam. So I have one
question: can you get gllam to estimate a fixed effects specification
rather than a random effects one?
--------------------------------------------------------------------------------
Your first model is conditional fixed effects, which wouldn't be matched by
what -gllamm- gave. Your second is the similar to what I had -gllamm- do,
but I believe -xtpoisson- uses a gamma distribution and not normal
distribution for the random effect by default, so it won't match exactly,
either. You might get even closer to what -gllamm- gave if you opted
for -xtpoisson , . . . re normal intmethod(aghermite) intepoints(8)-,
instead. Note that even then, there would be the -robust- option difference
between the two.
I'm not aware of -gllamm- being used to fit conditional fixed effects
models.
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/
