【作者】 韩阳；

【导师】 王德辉；

【作者基本信息】 吉林大学 ， 保险（专业学位）， 2015， 硕士

【摘要】 自从我国改革开放30多年以来,国内的金融市场不断向外界敞开怀抱,多种多样的金融产品不断被创新和改造.我国的金融市场竞争形势也更加激烈.如何准确的发现风险、度量风险和制定适应的风险决策,成为风控者们关注的核心问题.通常情况下,由于股票的日收益率之间的关系是非线性相关关系,恰好Copula理论可以反映股票日收益率之间的这种非线性相关关系,而且Copula函数可以结合单个股票的日收益率分布情况(边际分布)来表示出多个股票的整体分布情况(联合分布),故Copula理论成为金融市场中常用的用于风险度量的方法.在实际应用中,一个Copula模型所对应的边际分布的选取是解决问题的关键.因为GARCH-t模型能够很好地描述金融时间序列“尖峰厚尾”的特性,所以大多数研究股票的案例分析中采用GARCH-t模型来描述日收益率分布情况,但是,这种金融时间序列的分布情况通常不是对称的,而t分布具有对称性,基于此原因,本文创建了GARCH-偏f模型,该模型能有效克服上述的问题,很好的反映出边际分布的非对称性.杨湘豫和夏宇的研究中GARCH-t模型中参数的估计采取极大似然估计法,但是,在遇到目标函数没有极大值的情况,该种方法很大程度上不能得出最优解.本文采用的贝叶斯方法将待估参数看作是服从某种分布的随机变量,并且事先提供这个参数的先验分布,即人们对有关参数的主观经验信息.在参数估计过程中先验信息和样本信息同时被使用.贝叶斯估计方法有效克服了极大似然估计方法的缺陷.本文使用了WinBUGS软件有效的对模型中的参数进行估计.WinBUGS软件的运行以MCMC方法为基础,它将所有未知参数都看做随机变量,然后对此种类型的概率模型进行求解.论文首先介绍了GARCH模型、Copula理论和贝叶斯理论,然后运用贝叶斯估计理论建立Copula-GARCH-偏t模型,最后把汽车行业日收益率作为研究对象,运用贝叶斯估计方法得到模型的参数估计,继而得出日收益率实时的相关关系.研究结果表明一汽轿车与比亚迪汽车日收益率之间存在较强的正相关性.更多还原

【Abstract】 Since China’s reform and opening up 30 years, the domestic financial markets continue to open arms to the outside world, a wide variety of financial products continue to be innova-tion and transformation. China’s financial market competition more intense. How accurately identify risks, measure risk and to develop Risk adapt decisions become the core issue of concern who control the wind.Under normal circumstances, the relationship between daily stock returns is nonlinear relationship, just Copula theory can reflect this non-linear relationship between the rate of daily stock returns, and Copula function can be combined with daily yield distribution of individual stocks situation (marginal distribution) shows the overall distribution to multiple stock (joint distribution), so Copula theory into financial markets commonly used method for risk measurement. In practice, a Copula model corresponding to the marginal distribution Select the key to solving the problem. Because GARCH-t model can describe the charac-teristics of financial time series "fat tail", so most of the case studies used in the analysis of stock GARCH-t model to describe the distribution of daily returns, However, this distribu-tion is usually not symmetrical, and t distribution has symmetry, for this reason, this paper create GARCH-skewed t model effectively overcome the above problems, well reflect the asymmetric marginal distributions, and GARCH-t model to estimate the parameters of the majority to take maximum likelihood estimation method. When using maximum likelihood estimation method, the objective function encountered no great value, this method can not be drawn largely Best solution.Research GARCH-t model parameters estimated that most take before the maximum likelihood estimation method, however, in the case of the objective function encountered no great value, this method can not be drawn largely optimal solution. Bayesian methods used in this paper will be seen as a random variable estimation parameters obey a certain distri-bution, and provided in advance prior distribution of the parameters, namely people, priori information in the parameter estimation process for the subjective experience information on parameters and sample information are used simultaneously. Bayesian estimation methods effective customer the maximum likelihood estimation method defects.This article uses the WinBUGS software effectively in the model parameters estimated .WinBUGS software to run the MCMC method based on it all the unknown parameters are seen as random variables, then the probability of this type of model is solved.The article first introduces GARCH model, Copula theory and Bayesian theory and Bayesian estimation theory to establish Copula-GARCH-skewed t model, and finally the automotive industry daily yield for the study, the use of Bayesian estimation methods Model parameter estimation, and then come daily yield real-time correlation. The results show that there is a strong positive correlation between FAW and BYD daily return.更多还原