A Modified Confidence Set for the Structural Break Date in Linear Regression Models
A Modified Confidence Set for the Structural Break Date in Linear Regression Models
复制标题
线性回归模型中结构突破日期的修正置信集
DOI：
10.1080/00927872.2016.1178892
复制
发表时间：
2016
影响因子：
1.2
通讯作者：
Yohei Yamamoto
中科院分区：
文献类型：
--
作者：
眞野智行;津田照久;Teruhisa Tsuda;Teruhisa Tsuda;T. Tsuda;T. Tsuda;Teruhisa Tsuda;津田照久;直井道生;直井道生;直井道生;直井道生;直井道生;Yohei Yamamoto
Elliott和Müller（EM）（2007）提供了一种通过反转局部最佳检验统计量的变体来构建结构性突变日期置信集的方法。以前的研究表明，EM方法产生一组具有准确的覆盖率，即使是一个小的休息，但是，该集往往过于冗长。本研究提出一个简单的修改，以恢复他们的方法，通过长期方差估计。根据文献，我们提供了一个渐进的理由，改进的方法比原来的方法下的非局部渐近框架。蒙特卡罗模拟结果表明，改进的方法比EM方法获得更短的置信集，特别是当突变较大或进行HAC校正时。修改后的方法可能会表现出轻微的错误时，休息是小的覆盖率;然而，覆盖率是更稳定的比替代方法时，休息是大的。我们将我们的方法应用于20世纪80年代后日本通货膨胀数据的水平转移。
Elliott and Müller (EM) (2007) provide a method for constructing a confidence set for the structural break date by inverting a variant of the locally best test statistic. Previous studies have shown that the EM method produces a set with an accurate coverage ratio even for a small break; however, the set is often overly lengthy. This study proposes a simple modification to rehabilitate their method through the long-run variance estimation. Following the literature, we provide an asymptotic justification for the improvement of the modified method over the original method under a nonlocal asymptotic framework. A Monte Carlo simulation shows that the modified method achieves a shorter confidence set than the EM method, especially when the break is large or the HAC correction is conducted. The modified method may exhibit minor errors in the coverage rate when the break is small; however, the coverage is more stable than alternative methods when the break is large. We apply our method to a level shift in post-1980s Japanese inflation data.