中文摘要：

本文基于分析泛函中心极限定理在高频金融数据单位根检验中的特征与性质，从随机模拟的角度，探讨了有限样本情况下具有GARCH-GED误差项金融时序的ADF单位根检验统计量乙、乙的统计性质，着重研究了不同水平下的分位数、模型滞后阶的设定及GED分布参数的变动对其检验统计特性的影响。随机模拟结果显示，若数据生成模型设定为AR-GARCH-GED过程，常用的ADF单位根检验统计量存在不同的统计性质，并对出现这种现象的原因进行了探索。

英文摘要：

Based on analysis of the Functional Central Limit Theorem （FCLT） under high frequency financial time series data, the ADF unit root tests of time series with GARCH-GED error term are considered by Monte Carlo simulation under finite sample cases. Studies focus on simulating the quantiles of test statistics Zp and Z, under different size. The effects of lagged orders on model selection and parameter of GED on those quantiles also are investigated, Simulation results show that the empirical unit-root testing statistics have different statistical characteristics if the data generating processes are AR-GARCH-GED types processes, The reasons for those have been explored.