【作者】 李霞；

【导师】 王德辉；

【作者基本信息】 吉林大学 ， 保险硕士（专业学位）， 2017， 硕士

【摘要】 本文主要研究了条件异方差时变自回归模型参数的贝叶斯估计,在时变自回归模型参数的贝叶斯估计研究基础上,扰动项的方差服从ARCH(1)模型,对模型的参数不施加服从随机游走过程之类的限制,而是假设不同时刻其系数的先后状态具有一定的相关性,并且时间间隔距离越短它的相关性表现越强烈.由于时间的不可重复性,通常我们只能获得单个样本条,在这种情形下采用贝叶斯估计方法对参数进行估计.由贝叶斯公式得到参数的后验分布,用后验均值估计参数;对于复杂的后验分布要计算满条件分布,如果满条件分布不是已知分布,则不能对其直接抽样,应用逐分量的Metropolis-Hastings抽样方法进行抽样,用样本均值估计参数,从而得到参数的贝叶斯估计.本文通过分析一个简单的条件异方差时变模型,对其进行数值模拟来展示这个模型系数的估计效果.对我国GDP增长率的实例分析表明通过贝叶斯估计方法得到模型参数的估计值,能够很好地揭示实际问题变化过程中存在的内在规律.更多还原

【Abstract】 This paper mainly studied the conditional heteroscedasticity bayesian estimation of the time-varying autoregressive model,the time-varying autoregressive model parameters based on bayesian estimation and variance of the disturbance to obey the ARCH(1)model,the parameters of the model is not obey random walk process,such as restrictions,but assumes that different times its successively state has certain correlation coefficient,and the shorter the time interval distance its correlation performance is more intense.As a result of the time the unrepeatable,usually we can only get a single sample,in this case using bayesian esti-mation method is used to estimate the parameters.Parameters are obtained by the bayesian formula,the posterior distribution,with the posterior mean estimate parameters;For com-plex posterior distribution to compute the conditional distribution,if the conditions are not known distribution,are not the direct sampling,each component of Metropolis-Hastings sampling method to sample,the sample mean estimate parameters,the bayesian estimation of parameters is obtained.Based on the analysis of a simple conditional heteroscedasticity time-varying model,carry on the numerical simulation to show the model coefficient esti-mates of the effect.On China’s GDP growth rate of example analysis show that estimates of the model parameters is obtained by bayesian estimation method,is able to reveal the inherent law of practical problems exist in the process of change.更多还原