O Elias Krainski vai passar uns dias aqui em Curitiba e no período vai apresentar um curso sobre o INLA. A enfase será mais na motivaçãoe "entranhas" do INLA (e nao apenas no uso do pacote e exemplos de dados) A seguir um resumo, tb em www.leg.ufpr.br/seminarios todos interessados são bem vindos Local: LEG/UFPR Periodo: 09-13/06/2014 Horario: 10:00 - 12:00 ------------------------------------------------------------------------------- The main objective of this course is to look at the ideas behind INLA. INLA has a history which can be followed by looking bacl at the sequence of works by Prof. H. Rue's. We start on the MCMC algorithms, including single site and block updates and the Gaussian approximation as a proposal. Then, the INLA algorithm is used as one further step more wich makes sampling from the posterior no longer needed for most inference tasks. An working example is used to show, in details, how the algorithms work. At end we can see that any additive linear model with conditional independent likelihood can be easily included in the framework. If time allows, we will look at the INLA package and at some tricks implemented to deal with more complex models.