3
$\begingroup$

In statistics, how do you know if your count data is over dispersed after applying a glm, family =poisson to the data. I have 24 results in total, 12 with treatment, 12 without. There are 3 species, for each species 4 are with treatment and 4 without. So 8 results for each species, totaling 24 results.

Is there a standard number dispersion needs to be below? I don't know how this is decided in any kind of statistical model, including the one mentioned above.

Improve this question
$\endgroup$
2
  • $\begingroup$ Depends on the hypothetical distribution, e.g. if you have count data you think might be Poisson distributed, but then you find the variance is much greater than the mean, so it is overdispersed compared to the theoretical Poisson distribution. $\endgroup$
    – Andy W
    Commented Aug 26, 2017 at 1:48
  • $\begingroup$ Are you referring to overdispersion in GLMs or something else? Can you give some context for your question? $\endgroup$
    – Glen_b
    Commented Aug 26, 2017 at 6:26

0

Reset to default