SUMMARYThis paper is the natural sequel to Madruga ef al. (Environmetrics, 547-56 (1994)) where data from literature were analysed. Here, three sets of experimental data were obtained at IPEN (Energetic and Nuclear Research Institute) from blood cultures from three patients with basal cellular carcinoma (1 male, age 68, and 2 females, ages 47 and 66), four healthy young subjects (2 males, ages 24 and 27, and 2 females, ages 20 and 27) and two older healthy subjects (1 male, age 50, and 1 female, age 46). These cultures were exposed in vitro to ' %o radiati0.n ranging from 0 to 500cGy. The frequencies of cells with one or two nuclei, with or without micronuclei (MN), are the responses. Models for mono-and binucleated cells were obtained. As suggested in Madruga et al., models for mononucleated are simpler than for binucleated cells. The novelty is that, based on the observed frequencies of micronucleated cells, we can generate credible intervals for unknown doses to which an individual was exposed. The purpose of this paper is to obtain calibration models that will be helpful in establishing credible intervals for unknown exposure doses of a given agent once we have the observed responses. The agent in our case is radiation and the responses are frequencies of mono-and binucleated cells with none, one and two or more micronuclei.The underline probability distribution of the vector of these frequencies, yi = ( y a , yil, yi2), for each dose i, and for each type of cell (mono-and binucleated), is the trinomial distribution with parameter vector xi = ( R~, nil, 7ri2), where 7rio + 7ril + 7ri2 = 1. Recall that yi0 is the frequency of cells with no MN, yil with one, and yi2 with two or more MN.The Bayesian approach is used taking the conjugate Dirichlet distribution as the prior. The CCC 1 180-4009/96/030325-07
