偏正态空间自回归模型的贝叶斯估计  

Bayesian Estimation of the Skew-Normal Spatial Autoregression Models

作　　者：胡明星 句媛媛 戴琳[1] 吴刘仓[1] HU Mingxing;JU Yuanyuan;DAI Lin;WU Liucang(Faculty of Science,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]昆明理工大学理学院,云南昆明650500

出　　处：《昆明理工大学学报（自然科学版）》2021年第6期144-151,共8页Journal of Kunming University of Science and Technology(Natural Science)

基　　金：国家自然科学基金项目(11861041);昆明理工大学自然科学研究基金项目(KKSY201907003)。

摘　　要：空间自回归模型是空间计量经济学研究中的一个重要模型,主要用于刻画空间单元间的相关性.在空间自回归模型的现有研究中,大都假设响应变量服从正态分布,然而,实际的数据可能呈现出非正态的情况,此时,仍然在正态假设下作统计推断会获得不合理甚至错误的结论.基于响应变量服从偏正态分布的假设,研究偏正态空间自回归模型的贝叶斯估计.借助Gibbs抽样和MH算法相结合的MCMC算法讨论该模型的贝叶斯估计.数值模拟和实证分析表明:1)提出的偏正态空间自回归模型与传统模型相比,可以更好地拟合偏态数据;2)采用MCMC算法对模型进行贝叶斯估计,可以更精准地估计未知参数.研究结果显示:采用MCMC算法得到偏正态空间自回归模型未知参数的贝叶斯估计值更精准.Spatial autoregression models are important models in the spatial econometric research,which are mainly used to describe the correlation between spatial units.In the existing studies on spatial autoregression models,most researchers assume that the response variables follow the normal distribution.However,the real data may not be symmetric,so making statistical inference under the normal hypothesis will obtain unreasonable or even wrong conclusions.This paper assumes that the response variables follow the skew-normal distribution,and the Bayesian estimation of the skew-normal spatial autoregression models is discussed.A Markov chain Monte Carlo(MCMC)algorithm is developed to evaluate Bayesian estimation of the model by combining the Gibbs sampler and the MH algorithm.The numerical simulation and empirical analysis show that:1)Compared with the traditional model,the skew-normal spatial autoregression models can better fit the skew data;2)Bayesian estimation of the model can estimate the unknown parameters more accurately by using MCMC algorithm.The research results show that the Bayesian estimation of the unknown parameters of the skew-normal spatial autoregression models is estimated more accurately by using the MCMC algorithm.

关 键 词：偏正态分布 空间自回归模型 贝叶斯估计 MCMC算法 

分 类 号：O212.8[理学—概率论与数理统计]