近单位根过程脉冲响应函数的置信区间

本文提出一个构造近单位根自回归过程脉冲响应函数的置信区间的新方法。新方法首先修正自回归系数估计的偏误，然后利用标准自举方法构造脉冲响应函数的置信区间。蒙特卡罗模拟结果表明在小样本时新方法的表现要明显优于已有的方法。新方法的实际覆盖率能够一致地达到或超过名义置信水平。

Confidence Intervals for Univariate Impulse Responses with a Near Unit Root

This article proposes a new method for constructing confidence intervals for the impulse response function of a univariate time series with a near unit root.This method corrects the bias in the estimators of coefficients firstly,and then uses standard bootstrap method to construct confidence intervals.Monte Carlo simulation proves that the new method is better than other methods and it can achieve nominal confidence level consistently.