Inferência bayesiana não-paramétrica para elicitação da função de contabilidade

• Ano de defesa: 2014
• Autor(a) principal: Penha, Débora Luzia [UNESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Computação - Matematica; Contabilidade; Compressão de dados (Computação); Inferencia (Logica); Teoria bayesiana de decisão estatistica; Computer science - Mathematics
• Link de acesso: http://hdl.handle.net/11449/108645

The elicitation is a process that allows the incorporation of a prior information provided by an expert on some quantity, unknown and of interest, to the information from the experimental data. It can be used in many applied areas of knowledge, especially in situations where experimental data are not numerous because of the di culty or cost to obtain them. In Bayesian non-parametric approach, the density or the function of interest can be estimated without imposing any restrictive assumptions on its shape. Thus, the data allows determine the estimate of the interested function rather than conditioning it to a given parametric class functions. The aim of this paper is to apply the Bayesian nonparametric elicitation method of priors proposed by Oakley and O'Hagan (2007), Moala (2009) and Moala and O'Hagan (2010) to estimate the reliability function, considered completely unknown in its form, based only on the information provided by the expert.