Abstract: Two frequently used toxicity endpoints for evaluation of developmental effects are the proportion of deaths/resorptions and the proportion of malformations. Either a double beta-binomial model or a Dirichlet-trinomial model can be used to model the two developmental endpoints in the presence of intralitter correlation. Both models assume that the proportion of death/resorption and the proportion of malformation are two independent developmental effects. A quasi-likelihood model is proposed under both double beta-binomial and Dirichlet-trinomial variations. This approach allows for modeling a dependency between the two proportions. Two examples are given for illustration.
Key words and phrases: Beta-binomial, Dirichlet-trinomial, intralitter orrelation, quasi-likelihood.