閾値自己回帰モデルの生産者製品在庫率指数の変化率への応用とモデル選択 : ベイジアン・アプローチ

概要

論文の詳細を見る
In this paper we analyzed change rate of Japanese Index of Producer's Inventory Ratio of Finished Goods（Producer Goods For Mining and Manufacturing）by threshold autoregressive model. Firstly we consider the threshold autoregressive model in which there is one threshold（hereafter TAR1 model）. We use the test which is presented by Tsay（1989）for this data and compare varidity of the TAR1 model with auto regressive model（hereafter AR model）of same lag. In two cases, the validity of TAR1 model is shown. But we can’t compare validity of TAR1 model with that of AR model whose lag is different. Then we calculate DIC（Deviance Information Criterion）for both AR model and two TAR1 model. DIC suggest validity of TAR1 model and that most valid model is TAR1 model（p=4, d=2）. Secondly we consider the threshold autoregressive model in which there are two thresholds（hereafter TAR2 model）. We calculate DIC for TAR2 model and compare with those of TAR1 model and AR model. DIC suggest validity of TAR2model and that most valid model is TAR2model（p=1, d=1）.
2009-07-31