RE: st: Mixed effects model with zero-inflated negative binomial outcome for repeated measures data

[Robert Long <W.R.Long@leeds.ac.uk>]

Thanks again ! I have Skrondal. & Rabe-Hesketh in the office so I will check it today.

As for which parameter estimates are needed, I don't /need/ the NG dispersion parameter and the zero inflation parameter - only as a means for fitting ZINB as the outcome, but as far as I understand they must be estimated in order to fit ZINB. As for clustering, the data are repeated measures on around 100 subjects. The random intercept and slope variance are not of primary interest in themselves - interest is centered on the interactions between time and 2 covariates - but of course it would not be right to ignore the clustering by subject. I think it is possible to leave out the random slopes, but not the random intercepts.

For completeness, here is the model current specification again in R

glmmadmb(Y ~ time*X1 + time*X2 + (time | Subject), 
	data=final,family="nbinom2", zeroInflation=TRUE)

Best regards
Robert Long


-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of W Robert Long
Sent: 13 January 2013 16:05
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: Mixed effects model with zero-inflated negative binomial outcome for repeated measures data

Thank you.

I did, however, have the impression that -gllamm- does not support the negative binomial distribution ?

I have looked through the http://www.gllamm.org/examples.html page and I didn't see any mention of zero-inflation. I also looked through the review article http://www.bristol.ac.uk/cmm/learning/mmsoftware/reviewgllamm.pdf and didn't find it mentioned there either.....

Do you have any reference, or other information, about the possible collinearity problem ?  I assume this refers to correlation of random effects in the binomial probabilities part and the mixture-model part ? 
It would, of course, be ideal if it were possible to fix one of these REs to zero, and/or the correlation to zero, but to allow the model to estimate this correlation would also be OK !

Thanks again.

Robert Long


On 13/01/2013 00:05, JVerkuilen (Gmail) wrote:
> On Sat, Jan 12, 2013 at 1:43 PM, W Robert Long <W.R.Long@leeds.ac.uk> wrote:
>
>> Is it possible to fit this kind of model in Stata ? If so, could you 
>> provide some pointers ?
> I believe you could specify such a model in -gllamm- and seem to 
> recall that one of the examples on the web page compare Poisson, NB, 
> ZIP and ZINB, but if you look at http://www.gllamm.org/examples.html
> the examples are there for you.
>
> That said, I suspect that a ZINB with mixed effects would be a pretty 
> tough model to estimate well in nearly any package given that the NB 
> overdispersion parameter, overdispersion created by zero inflation, 
> and overdispersion created by clustering are all likely to be highly 
> collinear.
> *
> *   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/

*
*   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/