基于稳定分布的ARCH模型均值变点Subsampling检验

Subsampling Tests for the Mean Change-point in ARCH Models with Stable Distribution Innovations

讨论了基于稳定分布的ARCH模型的均值变点检验问题,其中特征指数k∈(1,2)。基于残量平方累积和统计量,利用Subsampling抽样方法确定渐近分布的临界值,从而避免特征指数k的估计。结果显示:蒙特卡罗数值模拟结果和实证分析充分说明了Subsampling抽样方法的可行性和有效性。因此,基于Subsampling的残量平方累积和检验对于稳定分布的ARCH模型均值变点检验仍不失为一种有效的方法。

The paper considers the problem of test for the mean change-point in ARCH models with innovations in the domain of attraction of a stable law with index.Based on the residuals CUSUM of squares test（RCUSQ）statistic,the subsampling method is proposed to determine the critical values of asymptotic distribution without estimating stable index k∈（1,2）.Mento Carlo simulation results and empirical study are performed to illustrate the validity and effective of this proposed method.In a word,the RCUSQ statistic based on subsampling is a fundamental tool to detect mean break in ARCH models with stable distribution innovations.