[R] Panel regression for zero-inflated count data with over-dispersion?

[Achim Zeileis]

On Fri, 13 Aug 2010, Steffen Krutzinna wrote:

> Hi,
> my first email, I hope I don't violate any rules.
> I want to run a panel (cross-section-time-series) regression with
> fixed effects. I have count data (patent citations) so a Poisson
> distribution applies. But I also want to control for over-dispersion
> and and excess zeros. I think I read all manuals of potentially
> helpful packages but I could not find a way to realise this. The
> closest I got was a fixed effects poisson regression using Yves
> Croissant's pglm package (panel generalized linear model). But
> over-dispersion and excess zeros are still neglected. Does anyone know
> a feasible way to run such a regression on R?

You can run a standard (i.e., cross-sectional) regression model and 
include a fixed effect (i.e., a factor indicating "id"). This is provided 
by zeroinfl() in the "pscl" package, see also the accompanying paper in 
the Journal of Statistical Software. Depending on how strong the 
autoregressive effects are, this might be a useful model.

There are a few other packages around which provide functionality for 
count data (see, among others, the "SocialSciences" and "Econometrics" 
task views on CRAN). But I don't think there is a dedicated package for 
panel analysis with overdispersion and excess zeros. (At least none that I 
know of.)

hth,
Z

> Thanks a lot in advance and thanks to all the help I got in the last
> years by reading the archived mailing lists.
>
> Steffen Krutzinna
> (Student @ Uni. Essex)
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>