论文部分内容阅读
现有随机波动(SV)模型依赖于参数条件分布形式假设,无法充分描述金融资产收益的偏态厚尾等典型特点,而非参数分布能够更全面地刻画这些特性。本文将SV模型和非参数分布相结合,构建一类半参数SV模型;同时在贝叶斯框架内,发展有效MCMC抽样解决模型的参数估计难问题,并利用对数预测尾部得分(LPTS)法分析模型的极端风险预测能力;最后以我国美元/人民币汇率市场为例,对半参数SV模型在收益特性刻画以及极端风险预测方面的实际效果进行了检验。
The existing stochastic volatility (SV) model relies on the assumption of parametric conditional distribution and can not fully describe the typical characteristics of skewed tail and tail of financial assets, while non-parametric distribution can describe these characteristics more fully. This paper combines SV model and non-parametric distribution to construct a class of semi-parametric SV model. At the same time, we develop an efficient MCMC sampling method to solve the problem of parameter estimation in Bayesian framework. By using Logistic Prediction Tail Score (LPTS) Finally, taking the China USD / RMB exchange rate market as an example, the author verifies the actual effect of the semi-parametric SV model in terms of income characterization and extreme risk prediction.