山形大学紀要(社会科学) 第36巻 第1号(2005)077-096

誤差項にt 分布を仮定した
閾値自己回帰モデルのベイズ推定

砂田 洋志

In this paper we introduce Threshold Auto Regressive Model, whose disturbance obey t distribution. We name this model t-TAR and show how to estimate the model by Bayesian MCMC method.
The difference of ordinary TAR model from t-TAR model is only distribution of disturbance. But when we estimate the model, the difference makes estimation difficult. We use complicated but efficient sampling method, rejection sampling chain.
In this paper we estimate 2 regime t-TAR model whose lag order is commonly 1, and confirm validity of the proposed method.




Bayesian Estimation of Threshold Auto Regressive Model,
whose disturbance obey t distribution

Hiroshi SUNADA




論文PDFファイル (1,326KB) はここをクリックしてください
Full Text PDF File (1,326KB)




目次・Contentsに戻る