[R] RE: zero-inflated count models (was polr problem solved)

[John Fox]

Dear Peter,

> -----Original Message-----

. . .
 
> The analyses were part of a paper I am writing, illustrating 
> that, when the DV is oddly distributed (the DV in question 
> was a count, with many 0's, and a long right tail) that the 
> 'usual' methods not only are wrong for statisically reasaons 
> (such as grossly violating model assumptions) but also give 
> bad results.
> 
> While this is widely known to statisticians, in the fields in 
> which I work, people sometimes analyze such variables using 
> either OLS regression (hence lm), or by categorizing the DV 
> into something like 0, 1, 2, more than 2 (hence the need for 
> polr). I also tried Poisson regression and negative binomial 
> regression (hence glm).  
> 

>From your description, it seems possible that there are too many zeros for a
Poisson or negative-binomial model. Since the focus of your paper is the
methodology, you might want to try a zero-inflated Poisson or
negative-binomial model. Though I haven't tried them, I'm aware of two
sources of R functions for zero-inflated count models -- zeroinfl(), from
Simon Jackman's web site <http://pscl.stanford.edu/content.html>, and gnlr()
in Jim Lindsey's gnlm package, which is not available on CRAN but at
<www.luc.ac.be/~jlindsey/rcode.html>.

I hope this helps,
 John

--------------------------------
John Fox
Department of Sociology
McMaster University
Hamilton, Ontario
Canada L8S 4M4
905-525-9140x23604
http://socserv.mcmaster.ca/jfox