脉冲噪声的干扰使随机信号出现厚尾性.对噪声项服从对称稳定分布的在线厚尾随机信号,提出了基于核加权的方差比率方法来连续监测其平稳性.给出了监测统计量在非平稳原假设下的极限分布,证明了此方法的一致性.为确定监测统计量的临界值并避免估计厚尾指数,提出Tbootstrap重抽样方法.最后通过模拟实验和分析两组实际数据说明了本文方法的有效性.
Impulse noise makes random signals occur heavy tails. For the online heavy tailed random signal with symmetrically distributed stable noise, we propose a kernel weighted variance ratio procedure to sequentially detect its stationarity. The asymptotic distribution of the monitoring statistic under nonstationary null hypothesis is derived, and its consistency is proved. In order to determine the critical values of the monitoring statistic and avoid the estimation of the tail index, we propose a bootstrap resampling method. Simulations and analysis of two groups of real data validate the proposed procedure.