Scott,
I don't know if this is relevant in your case, but you might try:
1. Switching between different optimization algorithms: technique(nr bhhh
dfp bfgs)
2. Rescale your variables.
"We also recommend that researchers with convergence
problems consider rescaling their variables. We have found
that simply rescaling the variables in a model so that all
their means are near 1 in absolute value will transform many
ill-conditioned problems to well-conditioned problems.
While this method is not the most general, its easy implementation
and high success rate make it an excellent place to start."
Drukker, D. and V. Wiggins. 2004. "Verifying the solution from a nonlinear
solver: a case study: comment." American Economic Review 94, 397-399.
Scott
> -----Original Message-----
> From: [email protected] [mailto:owner-
> [email protected]] On Behalf Of Scott Cunningham
> Sent: Wednesday, April 19, 2006 4:07 PM
> To: [email protected]
> Subject: st: poisson will not converge
>
> I am estimating both a linear fixed effects model and a POisson fixed
> effects model. The data is individual-level longitudinal data from
> the National Longitudinal Survey of Youth 1997. I want to control
> for individual, state, year and state*year fixed effects. I can do
> this in OLS using -xtreg-, but when I attempt to add the state fixed
> effects (let alone the state*year fixed effects) in the POisson
> model, it will not converge. I let it run for 10,000 iterations and
> it was clearly unable to finish. Since Poisson is not a
> computationally difficult likelihood to solve, I think something must
> be going on with the state fixed effects. How do I brainstorm my way
> out of this? What's causing this to happen?
*
* 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/
