Re: st: xtlogit: panel data transformation's recast to double makes model incomputable

[Jeph Herrin <stata@spandrel.net>]

To the other very good suggestions here, I'd add another. You don't say anything about your model, but if you are 
estimating an -fe- model and some (or even one) of the panels have all 0s or all 1s, the model will not converge. I have 
fit similar models to yours, but I usually have to drop all panels with no variation first.

You will also likely have problems if any of your `indvars' are panel level variables.

hope this helps,
Jeph



On 4/2/2013 8:28 AM, JVerkuilen (Gmail) wrote:
> On Tue, Apr 2, 2013 at 8:24 AM, Tom <tommedema@gmail.com> wrote:
>
> > Insofar noone has explained why the computation is unsuccessful. It'd
> > be especially helpful if someone could explain why less precision
> > would allow the maximizer to do it's job.
>
> Well I don't know but one direction to check is that that is a huge
> dataset. So try randomly subsampling down to, say 10% and see if that
> works. Maybe you're running into some weird memory or matrix size
> problem. Usually Stata would appropriately whine, but who knows.
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/