[SciPy-User] the skellam distribution

Ernest Adrogué eadrogue@gmx....
Tue Nov 10 16:47:05 CST 2009


 9/11/09 @ 11:58 (-0500), thus spake josef.pktd@gmail.com:
> from
> Bayesian analysis of the dierences of count data
> D. Karlis and I. Ntzoufras
> STATISTICS IN MEDICINE
> Statist. Med. 2006; 25:1885–1905
> 
> they have some funny application to soccer scores
> http://stat-athens.aueb.gr/~jbn/publications.htm

Yes, I have working on modelling soccer scores for some
time now, and have come to the conclusion that Poisson models
are doomed, because in essence football scores are not
Poisson distributed. If you are interested in these things
let me suggest you to concentrate your efforts on the negative
binomial model, no matter how tempting the Poisson model may be.

Cheers.
Ernest


More information about the SciPy-User mailing list