摘要
针对现有金融时间序列模型建模方法难以刻画模型参数的渐变性问题,利用贝叶斯分析方法构建贝叶斯厚尾SV模型。首先对反映波动性特征的厚尾金融随机波动模型(SV-T)进行贝叶斯分析,构造了基于Gibbs抽样的MCMC数值计算过程进行仿真分析,并利用DIC准则对SV-N模型和SV-T模型进行优劣比较。研究结果表明:在模拟我国股市的波动性方面,SV-T模型比SV-N模型更优,更能反应我国股市的尖峰厚尾的特性,并且证明了我国股市具有很强的波动持续性。
To solve the problem that the existing stochastic volatility model cannot describe the characteristics of parameters＇ time-changing, this paper establishes the Bayesian heavy-tailed volatility model. The paper firstly studies the model＇s statistical structure, chooses the parameter＇s prior distribution, designs a Markov chain Monte Carlo algorithm procedure with Gibbs sampler to carry out simulation analysis, and compares the SV-N model and SV-T model in the quality using the DIC criterion. The results indicate that, in modeling the volatility in the Chinese stock market, the SV-T model is superior to the SV-N model, which can better characterize the leptokurtic of stock returns. Furthermore, the results also prove that the Chinese stock market has high persistence of volatility.
出处
《运筹与管理》
CSCD
2007年第4期111-115,共5页
Operations Research and Management Science
基金
教育部新世纪优秀人才支持计划项目(NCET050704)
教育部人文社会科学规划项目(06JA910001)
湖南大学985工程项目
