On semi-parametric models in occult tumour experiments

Nonparametric statistical analyses of rodent carcinogenicity experiments involving occult tumours of intermediate lethality require numerous interim sacrifices. In recent papers, various semi-parametric analyses have been proposed which specify a parametric relationship between the death rate in tumour-bearing and tumour-free animals as a means of relaxing the requirement for multiple interim sacrifices. These semi-parametric methods involve different combinations of discrete and continuous time scales for the death and tumour-onset processes. The effect of combining different scales on this parametric relationship is studied further and illustrated via an example.