não disponível

In this work we define the medical diagnostic as an updating process of the diagnostic probabilities of a patient while clinical indicants are being obtained. Informations contained in the clinical data bank are also considered in the diagnostic. Part of the work is devoted to modellings of the diagnostic taking into account different types of experiments which are responsible for the generation of the data, in our case discrete. The modelling is done using a small set of restrictions, namely permutability and partial permutability. The theorem of de finetti guarantees a parametric representation of the clinical data as a mixture of multinomials, and the diagnostic becomes completely determined when we assume that the mixing function belongs to the dirichlet family. The choice of the family is very convenient mainly because the clinical data may be incomplete. The amounts of informatoins and evidences, given by data and indications obtained from the patient, are defined as a function of the probability distribution of the diagnostic. These quantities are used as descriptive statistics of the diagnostic. Some of them measure the capacity of the diagnostic of the patient, and others measure the weight of evidence in favour of hypotheses on the health state of the patient