周辺尤度の推定理論とその応用 : ギブス・サンプリング法、及びM=H法で推定した場合

概要

論文の詳細を見る
After we estimate several economic models by Bayesian method, we have to choice one model by Bayes Factor. Estimation of Bayes Factor is based on marginal likelihood, Then in this article, we introduce how to estimate marginal likelihood.　Gibbs sampling and Metropolis=Hastings method are familiar Bayesian estimation methods. When we estimate the parameters of the model by Gibbs sampling, we estimate marginal likelihood by Chib (1995). When we estimate the parameters by Metropolis=Hastings method, we estimate it by Chib=Jeliazkov (2001).　We introduce how to estimate the parameters of the model by Gibbs sampling, that is regression models with auto regressive error term. And we explain how to estimate the marginal likelihood. After we simulate the data from the model, we estimate both the parameters and marginal likelihood.　Then we introduce how to estimate the parameters of the model by Metropolis=Hastings method, that is auto regressive models with ARCH error term. And we explain how to estimate the marginal likelihood. After we simulate the data from the model, we estimate both the parameters and marginal likelihood.
2006-02-15