For non-normal data assumed to have distributions, such as the Poisson dist
ribution, which have an a priori dispersion parameter, there are two ways o
f modelling overdispersion: by a quasi-likelihood approach or with a random
-effect model. The two approaches yield different variance functions for th
e response, which may be distinguishable if adequate data are available. Th
e epilepsy data of Thall and Vail and the fabric data of Bissell are used t
o exemplify the ideas.