st: XTGEE in stata 10

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

st: XTGEE in stata 10

Narasimhan Sowmyanarayanan
Hi All:

I am working on some data in which the dependent variable is a count
number. The data is over-dispersed. So I am trying to work with xtgee.

xtgee [depvar] [independent var]  , family(nbinom) link(log)
corr(exchangeable) exposure(year)

I keep getting the error that says:

estimates diverging (missing predictions)
r(430);

I am using Stata 10.1. I noticed in one of the postings on the stata
website that this was a bug in stata 9 that has been fixed ? I also
tried to look at old postings on the stata listserv to see if any
solutions suggested there helps. None of them help me. Can someone
tell me what the problem can be and how I can solve this. Any insights
will be highly appreciated.

In particular, this happens if I add some non-linear terms and
interaction terms in the model. It does not converge. I tried building
the model one variable at a time. i can see that the moment I add the
non-linear xtgee stops working. I have used the same model in other
contexts too with non-linear terms such as OLS regression xtreg

Narasimhan
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
Reply | Threaded
Open this post in threaded view
|

st: RE: XTGEE in stata 10

Nick Cox
The evidence here to me implies that you are just trying to fit a model
that can't be fit to your data. With panel data any number of things
could be problematic.

Bear in mind that what to you is a problem -- Stata has "stopped
working" -- is to Stata a problem in that you asked an unanswerable
question.

This falls a long way short of a bug report that anyone could
investigate. Showing that the same data and the same model could be
fitted satisfactorily in other software would be an example of a
situation where StataCorp would want to see the details.

Nick
[hidden email]

Narasimhan Sowmyanarayanan

I am working on some data in which the dependent variable is a count
number. The data is over-dispersed. So I am trying to work with xtgee.

xtgee [depvar] [independent var]  , family(nbinom) link(log)
corr(exchangeable) exposure(year)

I keep getting the error that says:

estimates diverging (missing predictions)
r(430);

I am using Stata 10.1. I noticed in one of the postings on the stata
website that this was a bug in stata 9 that has been fixed ? I also
tried to look at old postings on the stata listserv to see if any
solutions suggested there helps. None of them help me. Can someone
tell me what the problem can be and how I can solve this. Any insights
will be highly appreciated.

In particular, this happens if I add some non-linear terms and
interaction terms in the model. It does not converge. I tried building
the model one variable at a time. i can see that the moment I add the
non-linear xtgee stops working. I have used the same model in other
contexts too with non-linear terms such as OLS regression xtreg


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/