Zero Inflated Poisson or Negative Binomial Regression model in Panel data

[Sagnik Bagchi]

Hi

I am using the Stata version 13.1. My data is count in nature and also it is a balanced panel. However, my data suffers from the problem of excess zeros and also suffers the problem of overdispersion. Having so I did for the Poisson Random Effects regression model and the Negative Binomial model.

I need to check for zero inflated poisson or negative binomial regression model. How does one do it in the current version of Stata .

Thanks,

Sagnik

[Carlo Lazzaro]

Sagnik:
I do hope that the following thread can help you out: http://www.stata.com/statalist/archi.../msg00731.html

[Sagnik Bagchi]

Thanks, Carlo. It would be great if you can kindly suggest me some papers regarding the same.

[Carlo Lazzaro]

Sagnik:
googling with the string: -zero inflated poisson panel AND Stata- gives back some promising entries.

[Marcos Almeida]

Hello Sagnik,

There was a similar question in the forum, and I share here the comments:http://www.stata.com/statalist/archi.../msg00829.html

I din't find any option in the Stata manual about zero-inflated negative binomial models in panel data.

I also checked it in a great book (Hilbe, Joseph M. Modeling Count Data. Cambridge University Press, 2014), particularly on the topic entitled "A word about longitudinal and clustered count models"), and I found no information on the availability of "mixing" panel data with zero-inflated models.

You know you can choose panel data (random-effects or fixed effects model) when dealing with count data, like when we type - xtnbreg - (http://www.stata.com/manuals13/xtxtnbreg.pdf)

As far as I'm concerned, and I may be wrong, there is only theoretical support for the use of zero-inflated count models in cross-sectional data.

This is the furthest I can go on this.

Hopefully a more "zero-inflated negative binomial" experienced member of the group will join in the discussion.


Best,

Marcos

[Sagnik Bagchi]

Thanks again, Carlo

[Sagnik Bagchi]

Thanks for all the help, Marcos. In corss section Zero Inflated models could be performed. In panel count data am not sure about how to do it in Stata.