基于贝叶斯ARFIMA-WRV模型高频数据长记忆性研究  

Study on Long Memory Property of High Frequency Data Based on Bayesian ARFIMA-WRV Model

作　　者：周树民[1] 陈健红 陈家清[1] ZHOU Shu-min;CHEN Jian-hong;CHEN Jia-qing(College of Science,Wuhan University of Technology,Wuhan 430070,China)

机构地区：[1]武汉理工大学理学院

出　　处：《数学的实践与认识》2019年第21期41-51,共11页Mathematics in Practice and Theory

基　　金：国家自然科学基金面上项目(81671633);中央高校基本科研业务费专项资金项目(2017IB011)

摘　　要：考虑到高频时间序列波动率的长记忆性问题,构建了赋权已实现波动分数整合自回归移动平均(ARFIMA-WRV)模型对其进行了研究.利用贝叶斯统计方法对模型做了相应的贝叶斯分析,并对我国中小板股市收益波动率的长记忆性特征进行了实证分析.实证结果表明我国中小板股市收益波动率存在长记忆性特征;采用消除日历效应影响的赋权已实现波动作为波动度量和贝叶斯参数估计方法,很大程度上提高了模型的参数精度.The long memory problem of high frequency time series is studied by constructing Bayesian weighted realized volatility autoregressive fractionally integrated moving average(ARFIMA-WRV) model.The Bayesian statistical method is used to make a corresponding Bayesian analysis of the model,and an empirical analysis is conducted on the long memory properties of the small and medium-sized board stock market volatility in our country.It is proved that the small and medium-sized board stock market volatility has long memory property.At the same time,the weighted realized volatility that can eliminate the effects of the calendar effect as volatility measurement and Bayesian parameter estimation are greatly improved the model’s parameter accuracy.

关 键 词：长记忆性 高频数据 ARFIMA-WRV模型 贝叶斯统计方法 

分 类 号：F83[经济管理—金融学]