摘要
通过分析随机波动模型的统计结构,推断了SV模型似然函数的具体形式,据此构造了模型参数的共轭先验分布.利用贝叶斯定理获得了相应的模型参数后验条件分布.同时,为了获得模型参数的贝叶斯估计及其置信区间,设计了基于Gibbs抽样的MCMC数值计算程序,并利用上海综合指数和深圳成分指数数据进行了建模实证分析,解决了参数随机条件下金融随机波动时间序列建模问题,提高了模型预报精度.
After exploring the statistical structure of the stochastic volatility model, its likelihood function＇s concrete form was derived, and its parameters＇ conjugate priors were constructed. Then, using the Bayesian theorem, the conditional posterior distributions were deducted. In order to obtain the parameters＇ Bayesian esti- mation value and their intervals, we designed a Markov chain Monte Carlo algorithm procedure with Gibbs sam- pler. Finally, using Shanghai stock＇ s comprehensive index and Shenzhen stock＇ s composition index, we gave an empirical example to illustrate how to use the proposed method. This paper provides a new approach to establish Bayesian models for financial stochastic volatility with random parameters.
出处
《湖南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2008年第12期88-92,共5页
Journal of Hunan University:Natural Sciences
基金
国家自然科学基金资助项目(70771038)
教育部新世纪优秀人才支持计划资助项目(NCET050704)
教育部人文社会科学规划资助项目(06JA910001)
