Bayesian analysis of stochastic volatility models with flexible tails
提出一种新的随机波动率模型，允许尾部形状由单独参数控制，从而灵活拟合金融数据中的厚尾和波动率聚类现象，并给出简单的马尔可夫链蒙特卡洛估计方法。
An alternative distributional assumption is proposed for the stochastic volatility model. This results in extremely flexible tail behaviour of the sampling distribution for the observables, as well as in the availability of a simple Markov Chain Monte Carlo strategy for posterior analysis. By allowing the tail behaviour to be determined by a separate parameter, we reserve the parameters of the volatility process to dictate the degree of volatility clustering. Treatment of a mean function is formally integrated in the analysis. Some empirical examples on both stock prices and exchange rates clearly indicate the presence of fat tails, in combination with high levels of volatility clustering. In addition, predictive distributions indicate a good fit with these typical financial data sets.