HAC法在平稳过程之间伪回归中的适用性研究

The Applicability of HAC Methods in Spurious Regression between Stationary Processes

通过蒙特卡罗模拟技术揭示各种HAC法在平稳过程伪回归中的适用性。研究发现,与核权函数HAC相比,预白化HAC法具有明显的优势;进一步的研究表明相对于被解释变量的持久性,解释变量的持久性对HAC的影响较大;当数据过程是高阶自回归过程时,在样本容量不是很大的情况下预白化方法的拒绝率会随着阶数增加而增大,只有在样本容量较大和BIC信息准则情况下预白化HAC的拒绝率才接近检验水平。

The paper reveals the applicability of HAC methods in spurious regression between stationary processes through Monte-Carlo simulation. The result shows that prewhitening HAC methods have obvious superiority to kernel-based estimators. Further study shows that the persistence of explanatory variables has greater effect on spurious regression than that of explained variable. When data processes are high-order autoregressive processes, the rejection rates of prewhitening HAC methods increase with the increasing order when sample size isn＇t very large. However, the rejection rates of prewhitening HAC methods is close to test size when sample size is very large and BIC information criterion is adopted.