基于APF算法的贝叶斯金融厚尾MSSV模型研究  

Bayesian Auxiliary Particle Filtering for Markov Switching Stochastic Volatility Models with Heavy-tails

作　　者：朱慧明[1] 郝立亚[1] 曾惠芳[1] 

机构地区：[1]湖南大学工商管理学院,湖南长沙410082

出　　处：《统计与信息论坛》2009年第9期3-9,42,共8页Journal of Statistics and Information

基　　金：国家自然科学基金项目<随机波动预测模型的贝叶斯分析及其在金融领域中研究>(70771038);教育部人文社会科学规划项目<时间序列计量经济模型的贝叶斯分析及其应用研究>(06JA910001)

摘　　要：针对金融时间序列普遍具有的波动聚集性和厚尾特征,将对风险管理尤为重要的一些极端点纳入模型之中,构建厚尾马尔科夫转移随机波动模型,采用带辅助变量的粒子滤波算法对波动和潜在状态进行预测,并估计模型参数。由于t分布与正态分布的特殊关系,通过选取不同自由度进行仿真分析,研究发现MSSV-t模型较一般MSSV模型对于消除波动持续性参数的高估问题更加有效。结合对中国上证综指股价波动的实证研究,证明了基于APF算法的MSSV-t模型在潜在波动状态的预测及突发事件的探测方面具有优良的性质,同时具备提高波动预测精度的能力。To capture changes in volatility which is caused by economic forces, market events and so on we studies the Markov switching stochastic volatility model with heavy - tails and its applications in this paper. However, because of the two hidden levels of the latent variable and the nonlinear sate space, it＇s hard to get the analytical evolution of the model. In the Bayesian perspective we applied the auxiliary particle filtering algorithm to filtering the hidden states and estimated the parameters. We focused on the effect of tails＇ heaviness and got the conclusion that the APF algorithm performed more efficient when the MSSV model with heavier tails. The model was also applied to a real financial time series： the Shanghai Composite Index and showed the ability to capture the different volatility regimes and to improve the estimation precise.

关 键 词：时间序列分析 贝叶斯推断 MSSV模型 APF 仿真 

分 类 号：F064.1[经济管理—政治经济学] O212.8[理学—概率论与数理统计]