统计研究 ›› 2012, Vol. 29 ›› Issue (4): 98-104.

• 论文 • 上一篇    下一篇

时间趋势平稳性检验的带宽选择及其影响

周少甫 左秀霞   

  • 出版日期:2012-04-15 发布日期:2012-04-17

Bandwidth Selection of Time Trend Stationarity Test and Its Effects

Zhou Shaofu & Zuo Xiuxia   

  • Online:2012-04-15 Published:2012-04-17

摘要: 本文将基于最小化两类错误的概率选择带宽的思想引入KPSS检验。文章通过蒙特卡罗仿真实验,对基于最小化均方误差和最小化两类错误概率的KPSS检验的有限样本性质进行了比较;并对中国季度实际GDP对数序列的平稳性进行了检验。结果表明,基于最小化两类错误概率的检验比基于最小化均方误差的检验有更大的水平扭曲,但前者也有更大的检验势;并且在误差项自回归系数小于等于0.5的大多数情形下,基于最小化两类错误概率的检验有更好的有限样本性质;中国季度实际GDP对数序列是带漂移的单位根过程。

关键词: KPSS检验, 长期方差, 两类错误

Abstract: This paper introduces the idea of determining kernel function’s bandwidth by minimizing the probability of two types of errors into KPSS test. In this paper, the Finite Sample Properties based on minimizing Mean Square Error and on minimizing the probability of two types of errors are compared by Monte Carlo Simulation. And the Stationarity of the logarithmic series of Chinese Quarterly Real GDP is tested. The results show that, the size distortion of the test based on minimizing the probability of two types of errors is larger than the one based on minimizing Mean Square Error, the test’s power, however, is also larger. And in most cases, when error term’s autoregressive coefficient is less than or equal to 0.5, the test based on minimizing the probability of two types of errors has better Finite Sample Properties. And the logarithmic series of Chinese Quarterly Real GDP is unit root process with drift.

Key words: KPSS Test, Long Run Variance, Two Types of Errors