Re: st: RE: RE: convergence problems with zinb

[Partha Deb <partha.deb@hunter.cuny.edu>]

Hi Margaret,

Given this output, you may want to consider another source of your issues. 
The output suggests that collinearities or near collinearities between your 
regressors, or "perfect" predictors in the inflate part might be the source of 
your convergence problems.  Do you have indicators for rare conditions?  Could 
there be other collinearities between conditions?  Someone already suggested 
starting with a pared down model - I would definitely try that.  In addition, 
it might be well worth seeing what happens if you remove chronic and/or 
asthbronc from your specifications.

Finally, I've observed convergence problems using nb-based models if there 
isn't evidence of overdispersion.  What is the value of alpha?

Hope these help.

Partha


Holland, Margaret wrote:

> Thanks for the helpful suggestions.
>
> I've tried a few of the different algorithms, with and without the
> "difficult" option and I'm getting closer (I think). I don't get any
> error messages, it looks like it converges, and I get coefficient
> estimates, but some of the standard errors are missing (or huge):
>
> -------------+------------------------------------------------------
> inflate      |
>      chronic |  -11.51164        .      .     .          .         .
>   asthmbronc |  -26.67018 193706.5  -0.00 1.000  -379684.4  379631.1
>
>
> I'm interpreting this as there was a problem in the estimation, but for
> some reason it didn't trigger an error message. Is this an appropriate?
> I'll continue working to refine the model based on that assumption, but
> was curious if there was anything else I should know.
>
> I have not yet taken the time to explore potential problems with my
> imputed data, but I suspect the problem is more in the model itself
> because I have seen the same problem in my raw (not imputed) data.
> However, I will take a look as suggested, just in case there is
> something unusual. I should probably have done that by now anyway...
>
> Thanks again for the help on this,
> Maggie
>
>
> -----Original Message-----
> Sent: Wednesday, August 13, 2008 8:04 PM
> Subject: st: RE: convergence problems with zinb
>
> Hi Margaret,
>
> You might like to look at the various maximize options that you can
> tinker with.  
> Type - help maximize
>
> There are four different algorithms used by ml and Stata follows a rule
> for stepping though these when fitting a model.  There is an option
> called - difficult - which tells the ml program to use a different
> stepping rule.  You could try this.
>
> Any algorithm used to maximize the log-likelihood has to start with some
> initial coefficient values and sometimes these can be near a flat or
> concave region of the likelihood function.  Consequently, the fitting
> algorithm will just wander around this region, unable to get out.
> Changing the initial values might therefore improve the performance of
> the algorithm.
>
> The option - trace - will give you the current coefficient estimates at
> each iteration.  This might give you a clue as to what's going on. 
> I don't know how many variables you have in your model, but you could
> try fitting a separate model for each variable thus obtaining a
> coefficient estimate for each variable and then use these as your
> initial estimates in the full model.
> 	
> Kieran
>
>
> *
> *   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/
--
Partha Deb
Department of Economics
Hunter College
ph:  (212) 772-5435
fax: (212) 772-5398
http://urban.hunter.cuny.edu/~deb/

Emancipate yourselves from mental slavery
None but ourselves can free our minds.
	- Bob Marley

*
*   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/