Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Nick Cox <njcoxstata@gmail.com> |
To | statalist@hsphsun2.harvard.edu |
Subject | Re: st: control a variable in stata |
Date | Sat, 21 Apr 2012 21:29:24 +0100 |
I agree with the main message, but as so often happens, you are 1. stressing secondary assumptions, namely assumptions about the stochastic part of a model 2. in some danger of confusing marginal distributions and conditional distributions. -poisson- does not depend on the marginal distribution of the response being Poisson any more than regression depends on being Gaussian. Nick On Sat, Apr 21, 2012 at 8:33 PM, Nora Reich <nhmreich@googlemail.com> wrote: > As far as I know, -poisson- is for skewed distributions, and salary > distribution in general often fulfills this requirement, but salary of > NBA players might show a different distribution. The distribution can > be checked, e.g. with the command > > -histogram salary- > > (for more information type -help histogram-). > > Andy, I would compare the assumptions and requirements of different > estimation strategies (OLS, poisson) and find out which fits better > with the data. > > I would also check which estimations strategies are used by similar > papers, and why. > > Apart from -poisson- and -regress-, I cannot think of any at the > moment that have to be considered for your salary-estimations. > However, if there was something like a "minimum wage" for NBA players, > i.e. salary is censored, -tobit- would be an alternative. > * * 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/