Document type
Academic documents
Document subtype
Master's dissertation
Title
Cointegração, modelos VAR e BVAR: Estudo comparativo entre a Abordagem Clássica e Bayesiana no contexto dos Mercados Financeiros Europeus
Participants in the publication
Filipe R. Ramos (Author)
Dep. Estatística e Investigação Operacional
Diana E. Aldea Mendes (Adviser)
Summary
Given the development of an econometric modeling study, the main objective of this work consists in the comparative analysis of results obtained in the estimation of a cointegrated vector autoregressive model and its cointegration rank, from the point of view of the assumptions of the classical (VAR) and Bayesian (BVAR) approaches. For this purpose a sample of variables of stock indices from six countries of the Euro area, were taken into account. Preceding to the practical implementation of the econometric models, we present a theoretical framework that aims to provide, in a gradual manner, information deemed necessary for the understanding of the econometric models in question. Thus, besides the inherent aspects of the classical approach, the Markov Chains Monte Carlo (MCMC) simulation methods are analyzed, by presenting in particular the Metropolis-Hastings algorithm and Gibbs sampling method, whose foundations were discussed based on Markov Chains theory. These methods, based on iterative stochastic simulation form the basis of bayesian inference, relying on the knowledge of the posterior distribution of parameters and on the possibility of constructing exact confidence intervals to estimate the considered parameters. Precisely, since the main tool of the implementation of this approach consists the numerical integration (in the parameter space) of the posterior fdp, the option to summarize the information described in the posteriori distribution are based on the MCMC methods. The advantages in the implementation of Bayesian techniques have been clearly identified in further studies (observe the quite common appeal to them) hence previously intractable problems have been resolved and more flexible models were introduced with great success.
Editor(s)
Faculdade de Ciências - Universidade de Lisboa
Date of Submisson/Request
2011
Date of Publication
2012
Institution
INSTITUTO SUPERIOR DE CIÊNCIAS DO TRABALHO E DA EMPRESA
Publication Identifiers
Document Identifiers
DOI -
https://doi.org/10.13140/RG.2.2.33358.79684
URL -
http://hdl.handle.net/10451/8822
Keywords
Inferência Bayesiana
Cadeias de Markov
Simulação MCMC
Cointegração
Modelos VAR
Modelos BVAR