Inferência bayesiana para o tamanho de uma população fechada com erros de registros de dados amostrais

Abstract

In this dissertation we determine maximum likelihood and bayesian estimates of the size of a closed population, from two lists of data of elements of the population. It has been supposed that the registers of the individual information in the lists are capable of mismatches and, with relation to the bayesian method, the prioris distributions are noninformative and they have maximum enthropy for the parameters. We also present the bayesian model, witch has considered the numbers elements of the two lists as a latent variable. We compare these models through examples with simulated and real data.