GARCH-M模型的MCMC分析及应用

发布时间：2018-02-28 16:27

本文关键词： GARCH-M 模型 MCMC 方法 M-H 抽样　出处：《华南理工大学》2012年硕士论文　论文类型：学位论文

【摘要】：GARCH-M模型在其均值方程中引入了条件方差项，从而使得该模型在刻画风险报酬、风险溢价等方面具有广阔的应用空间。由于MCMC方法能把一些复杂的高维问题转化为一序列简单的低维问题，，同时不要求似然函数和参数先验分布具有共扼结构，因此通过构造合理的转移核来得到GARCH-M模型所需的后验分布，进而可以进行相关的贝叶斯分析及其模型预测。本文通过MCMC方法计算后验分布的数值积分来实现对GARCH-M模型的参数估计。采用基于M-H抽样的MCMC方法探讨了模型参数的贝叶斯估计，通过MCMC方法的数值计算，给出预测分布，并进行数据模拟。最后给出GARCH-M模型在黄金市场的应用、得出黄金现货收益率尖峰厚尾性、聚集性等相关性质，以及预测黄金现货价格的走势。
[Abstract]:The conditional variance term is introduced into the mean equation of GARCH-M model, which makes the model depict the return of risk. Because the MCMC method can transform some complex high-dimensional problems into a series of simple low-dimensional problems, the likelihood function and the prior distribution of parameters are not required to have a conjugate structure. Therefore, the posteriori distribution of GARCH-M model can be obtained by constructing a reasonable transfer kernel, and then the relevant Bayesian analysis and model prediction can be carried out. In this paper, the parameter estimation of GARCH-M model is realized by calculating the numerical integral of posterior distribution by MCMC method. The Bayesian estimation of model parameters is discussed by using MCMC method based on M-H sampling, and the predicted distribution is given by the numerical calculation of MCMC method. Finally, the application of GARCH-M model in gold market is given, and the related properties of gold spot yield, such as peak, thick tail, aggregation and so on, are obtained, and the trend of gold spot price is forecasted.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F224;F831.54;F416.32

【参考文献】