混合贝塔分布随机波动模型及其贝叶斯分析

为了更准确地揭示金融资产收益率数据的真实数据生成过程,提出了基于混合贝塔分布的随机波动模型,讨论了混合贝塔分布随机波动模型的贝叶斯估计方法,并给出了一种Gibbs抽样算法。以上证A股综指简单收益率为例,分别建立了基于正态分布和混合贝塔分布的随机波动模型,研究表明,基于混合贝塔分布的随机波动模型更准确地描述了样本数据的真实数据生成过程,而正态分布的随机波动模型将高峰厚尾等现象归结为波动冲击,从而低估了收益率的平均波动水平,高估了波动的持续性和波动的冲击扰动。

Stochastic Volatility Model Based on Mixed Beta Distribution and Their Bayesian Analysis

In order to more accurately describe the true data generating process of financial assets yield data, we first proposed a stochastic volatility model based on mixed beta distribution （SV--M）, then discussed Bayesian estimation method of the SV--M model, and gave the Gibbs sampling algorithm. Finally, taking the Shanghai A shares KLCI simple rate of return as an example, we established SV--M model and SV--N model（the stochastic volatility model based on the normal distribution）, and made a comparative analysis, the empirical analysis suggested that SV--M model more accurately described the real data generation process of the sample data; SV--N model attributed peak thick tail phenomenon to the impact of volatility, so that underestimated the average yield volatility levels, overestimated the fluctuations and the impact of fluctuations in disturbance.