周期异方差时间序列中季节单位根的Gauss检验

Gaussian Tests for Seasonal Unit Roots in Periodical Heteroscedastic Time Series

为检验带异方差的季节时间序列中的单位根,提出了基于Cauchy估计的Zc统计量.在原假设下得到该检验统计量的极限分布服从于标准正态分布,并与季节周期d和误差项的周期异方差无关.Monte Carlo模拟计算表明当模型中自回归系数接近1时,用该方法得到的估计值比普通的最小二乘方法得到的估计值误差小.计算结果和实例分析结果表明了用该方法检验季节单位根的简便性和有效性.

The paper propose test statistics based on Caucby estimation for unit root with periodical heteroscedasfic shocks in seasonal time series. Under null hypothesis, the limit distributions of statistics are standard normal regardless of the period of seasonality and periodical heteroscedasticity of shocks. By Monte Carlo simulation, it is obtained that the estimators have smaller errors by this method than LSE when the autoregressive coefficient is close to 1. Simulations and a real example show the convenience and validity of this test method.