Binomial data in presence of overdispersion

Abstract

In many applications of binomial data independent observations in different groups can have positive correlation within groups so that the variance is larger than under independence (overdispersion) Amongst the various models that can be fitted to deal with the phenomenon of over-dispersion, the use of quasi-likelihood functions is now days a well established one. In this paper we consider a method for the estimate of the dispersion parameter based on a procedure proposed by Williams (1982) that can be applied to the case of grouped data. As an application, a set of data on an experiment to examine the infestation caused by an insect on the olive tree is analysed.