Analysis of proportions of affected foetuses in teratological experiments

This paper is concerned with the analysis of proportions affected when an increasing dose of a compound is applied to a group of laboratory animals. Several sets of data, including a set from a teratological experiment at the Shell Toxicology Laboratory, Sittingbourne, are analysed and some simulations are performed. Among the distributional models, the beta-binomial model is, in general, found to be the most sensitive to departures from the binomial. For testing the equality of two proportions, a comparison is made between the pseudo-t test, based on the jackknife method, and the likelihood ratio test, based on the beta-binomial model. From the limited comparison, no definite advantage of one approach over the other has been found; at the 5% level of significance both approaches lead to similar conclusions.