基于Gibbs抽样门限自回归模型的参数估计  

Parameter Estimation Based on Gibbs Sampling Threshold Autoregressive Model

作　　者：蒋捷 郑月晨 周浩 张慧增[1] JIANG Jie;ZHENG Yuechen;ZHOU Hao;ZHANG Huizeng(School of Science,Hangzhou Normal University,Hangzhou 311121,China)

机构地区：[1]杭州师范大学理学院,浙江杭州311121

出　　处：《杭州师范大学学报（自然科学版）》2020年第4期410-414,共5页Journal of Hangzhou Normal University(Natural Science Edition)

基　　金：国家自然科学基金青年项目(11901145).

摘　　要：基于贝叶斯统计推断以及模型随机项的基本假设,通过门限自回归(TAR)模型各参数与总体的共轭先验分布,利用蒙特卡洛模拟(MCMC)算法和Gibbs抽样从各参数的后验分布抽样,用后验均值来估计TAR模型的待估参数.通过模拟实验进一步验证基于贝叶斯统计推断在TAR模型参数估计中的有效性.Basing on Bayesian statistical inference and the basic assumptions of model random terms,the conjugate prior distribution of each parameter and population of the threshold autoregressive model TAR is used to estimate the undetermined parameters of TAR model with posterior mean value by using the MCMC algorithm and Gibbs sampling from the posterior distribution of each parameter.The simulation experiments prove that the validity of Bayesian statistical inference in the parameter estimation of threshold autoregressive model TAR.

关 键 词：贝叶斯统计推断 TAR模型 共轭先验 蒙特卡洛模拟 GIBBS抽样 

分 类 号：O212.8[理学—概率论与数理统计]