Methods For Nonstationary And Nonlinear Models In Economic Time Series And Their Applications

本项目涉及非平稳和非线性时间序列回归模型的理论和应用。我们的研究可以分为三类：(1)非平稳和非线性性，(2)非平稳，(3)回归理论，这是(1)和(2)研究的基础。关于(1)，Maekawa和Tee提出了当时间序列有单位根时，估计时间序列结构变化时间点的方法。我们的研究结果已在澳大利亚堪培拉举行的日本统计协会年会和MODSIM2001年会上公布。B)关于(2)Maekawa和Hisamatsu通过渐近理论和蒙特卡罗模拟研究了具有非平稳回归变量的SUR模型，发现Zellner的两个估计量在平稳情形下优于OLS估计量。这一结果发表在《应用计量经济学和统计推断手册》(2002年，Marcel Dekker，Inc.)上。Maekawa和Zongle He证明了通常的格兰杰因果关系的F检验经常检测到两个独立且不相关的时间序列之间的伪因果关系，其中一个或两个是随机游走(S)。这一结果发表在2001年的《经济学快报》上。(C)关于(3)Kurata研究了SUR模型和一般线性回归模型中GLS估计量的统计性质。福口研究了时间序列回归中的次抽样方法。此外，我们邀请了首尔大学的J.Park教授听取了非平稳和非线性时间序列的最新发展。前川到访新加坡国立大学，与谢育德教授会面，讨论金融时间序列中的非平稳和非线性TOME序列回归模型。

This project concerns with theory and application of non-stationary and non-linear time series regression models. Our studies can be grouped in the three categories : Category (1) deals with both non-stationary and non-linearlity, (2) deals with non-stationary, (3) theory of regression which is the base of studies of (1) and (2). We obtained the following results.A) Concerning (1) Maekawa and Tee proposed a method of estimating a point of time of structural change in a time series when a time series has a unit root. Our result was presented at the annual meeting of Japan statistical association and MODSIM2001 held in Canberra, Australia and conferences. Our method is found to be better than other existing methods in the literature on this problem.B) Concerning (2) Maekawa and Hisamatsu studied SUR model with non-stationary regressors by asymptotic theory and Monte Carlo simulation and found that Zellner's two estimators were superior to OLS estimator as in the stationary case. This result is published in Handbook of applied econometrics and statistical inference (2002, Marcel Dekker, Inc.). Maekawa and Zongle He showed that usual F test for Granger's causality often detected spurious causality between two independent and irrelevant time series one of or both of which is/or random walk(s). This results was published in Economics letters, 2001.(C) Concerning (3) Kurata studied statistical properties of the GLS estimator in SUR model and a general linear regression model. Fukuchi studied subsampling method in time series regression.In addition we invited Professor J. Park at Seoul University to hear the recent development on non-stationary and non-linear time series. Maekawa visited the National University of Singapore to meet Professor Y. Tse to discuss non-stationary and non-linear tome series regression models in financial time series.