基于沪深300高频数据的非参数跳跃识别方法比较
首发时间:2015-09-11
摘要:针对现有非参数跳跃识别方法繁多,但对其在中国股票市场的适用性尚无定论的问题,以沪深300指数高频价格为实证数据,比较各种跳跃检验方法应用于中国股票市场的优劣性。除了对各种跳跃检验方法检验出的跳跃个数进行比较外,着重于将跳跃方差和连续路径方差用于HAR-RV-CJ模型的样本内拟合结果与样本外预测结果的比较。对样本外预测结果进行比较时将其分为22个周期分别进行SPA检验。通过对沪深300指数数据进行研究,我们发现样本内拟合结果差别不大,系数显著度水平以及拟合优度都在可接受范围内;而从样本外预测结果来看,Z_Log-Med检验表现最好,无论是在整个预测区间评估还是对预测区间分段评估时都具有明显的预测性能优势。但是,同时也可以发现这种优势在日预测中最显著,随着预测期的延长,优势越来越小,也就是说,在进行长期预测时,各种跳跃检验方法差别不大。
关键词: 非参数跳跃识别 高频价格 HAR-RV-CJ模型 SPA检验 波动率预测
For information in English, please click here
Comparison of non-parametric jump tests based on high-frequency data of Chinese CSI 300 index
Abstract:Various non-parametric jump tests have been designed, while their applicability in Chinese stock market has not been deeply studied. Thus this article empirically compares various jump tests in Chinese stock market based on high-frequency prices of Chinese CSI 300 index. In addition to comparing the number of identified jumps , this paper also compares the in-sample fit and out-of-sample forecast results of the HAR-RV-CJ model which separates the contribution of jump variance and continuous variance. The out-of-sample prediction period is divided into 22 sub-periods and the SPA test is used for comparison in each sub-period. Using Chinese CSI 300 index as empirical data, we find that various jump tests lead to similar in-sample fit results, with coefficient significance level and goodness of fit all within acceptable ranges. As for the out-of-sample forecast results, Z_Log-Med test has obvious advantages considering both the total forecast period and each sub-period. However, such advantage is most obvious for daily forecast, and gets less obvious as the forecast period prolongs.
Keywords: non-parametric jump test high-frequency prices HAR-RV-CJ model SPA test volatility forecasting
论文图表:
引用
No.4653959894747144****
同行评议
共计0人参与
勘误表
基于沪深300高频数据的非参数跳跃识别方法比较
评论
全部评论0/1000