Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | David Hoaglin <dchoaglin@gmail.com> |
To | statalist@hsphsun2.harvard.edu |
Subject | Re: st: Model for Poisson-shaped distribution but with non-count data |
Date | Tue, 6 Dec 2011 07:33:37 -0500 |
In a quick look at the blog that Nick mentioned, I did not see any mention of the fact that the Poisson distribution is discrete. In the limit (as the mean of the distribution becomes large), that matters less, but one would need to view the possible data values as discrete. Some of the equations in the blog are not quite correct. For example, since Poisson regression is a form of generalized linear model, the linear predictor is fitted to log(E(y)), rather than to log(y). The random component of the GLM is a Poisson distribution. David Hoaglin On Tue, Dec 6, 2011 at 4:00 AM, Nick Cox <njcoxstata@gmail.com> wrote: > -gammafit- (SSC) has been available for some years, but random > intercepts are fancier than it does. > > However, I am more concerned with two dogmas surfacing here without > little or no foundation, that > > 1. Poisson models are for counts only > > 2. You choose models based on the marginal distribution of the > response or outcome variable. > > See http://blog.stata.com/2011/08/22/use-poisson-rather-than-regress-tell-a-friend/ > for an excellent exposition that makes no such assumption on Poisson. > On the evidence here I would still try out -poisson- or a related > command. > > I don't know where #2 comes from. Every decent modelling text > explains that assumptions are about conditional distributions, and not > that important even then. > > Nick * * 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/