Workshop on Probabilistic and Statistical Methods - Apresentação de comunicação oral
Meu trabalho sobre ABC usando GEL e SMM foi aprovado para comunicação oral no Workshop on Probabilistic and Statistical Methods, que acontecerá na semana que vem no ICMC-Usp em São Carlos.
Eu iria nesse excelente workshop que qualquer forma, mas é sempre uma honra apresentar um trabalho.
Approximate Bayesian Computing via Generalized Empirical Likelihood and Method of Simulated Moments.
Márcio Poletti Laurini
FEA-RP USP
Abstract
An important recent literature discusses the use of Bayesian methods using numerical approximations which are likelihood free (Approximate Bayesian Computation - ABC). These methods allow inferences on issues
hitherto untractable due to the complexity of the evaluation of the likelihood function. We discuss how to formulate ABC methods using Generalized Empirical Likelihood, extending the work of Mergensen et al
(2012). These methods allow a formulation computationally more efficient than the usual ABC methods, and can be used even when it is not possible to calculate the moment conditions analytically through the use of
simulated moment conditions. We apply these methods in the estimation of Stochastic Volatility Models.