To all allstat members,
I'd appreciate some suggestions for the following problem:
I try to fit a poisson distribution to a count variable,
abundance of a certain species. For common species the poisson
distribution seems not apt, but for the rarer species it might
work. I'd like to assess the goodness of fit of the poisson
distribution.
I have 5 replicates of the count variable which complicates the
situation. The sum of 5 poisson RV should also be poisson, but
if I want to calculate a X^2 goodness of fit statistic for the
sum, then there are many "holes" in the computation in the sense
that many theoretical count values have never been observed.
Use of a kolmogorov smirnoff statistic neither seems appropriate,
since this requires that the null distribution is continous, and
that is not the case.
Any suggestions for how to assess goodness of fit ?
Best regards, Jan Graffelman
---------------------------------------------------------
| Jan Graffelman | internet: [log in to unmask] |
| Universitat Pompeu Fabra | tel: 34-93-5421621 |
| Ramon Trias Fargas 25-27 | fax: 34-93-5421746 |
| 08005 Barcelona, Spain | |
---------------------------------------------------------
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|