In glms we can use a quassipoison fudge factor to account for over dispersion in our poisson models.

In glmms we can add an individual level random effect (e.g. id) for each row in data.frame to account for over dispersion. i.e.

glmer(y ~ x + (1|group) + (1|id), family = poisson)

Can someone give me a feel for how this individual level random effect deals with over dispersion?

  • $\begingroup$ have you read the material about overdispersion and observation-level random effects at http://glmm.wikidot.com/faq ? $\endgroup$
    – Ben Bolker
    Nov 20, 2014 at 0:29

1 Answer 1


If your concern is for whether this method is adequate, a recent simulation paper on PeerJ "Using observation-level random effects to model overdispersion in count data in ecology and evolution" indicates that individual-level random effects works well but do have limitations. In the abstract the author states: "This work suggests use of observation-level random effects provides a simple and robust means to account for overdispersion in count data, but also that their ability to minimise bias is not uniform across all types of overdispersion and must be applied judiciously."

I believe Bolker does some comparisons of different ways to deal with overdispersion here and works through the classic Grouse Tick data of Elston et al.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.