半参数空间滞后分位数回归模型的贝叶斯估计及应用  

Bayesian estimation and application of semiparametric spatial lag quantile regression model

作　　者：方丽婷 李坤明 FANG Liting;LI Kunming(School of Economics and Management,Fuzhou University,Fuzhou 350108,China;College of Economics and Management,Fujian Agriculture and Forestry University,Fuzhou 350002,China)

机构地区：[1]福州大学经济与管理学院,福州350108 [2]福建农林大学经济与管理学院,福州350002

出　　处：《系统工程理论与实践》2024年第10期3346-3361,共16页Systems Engineering-Theory & Practice

基　　金：国家自然科学基金(72003034);国家社会科学基金(22BTJ026);福建省自然科学基金(2021J01113);福建农林大学科技创新专项基金(KFB22106XA)。

摘　　要：文章提出一类半参数空间滞后分位数回归模型,该模型能够同时考察因变量的空间相关性和影响机制的部分非线性,且可以在任意分位点对响应函数分别建模.其次,文章构建了模型的贝叶斯估计方法,在贝叶斯理论框架的构建中,文章利用多项式样条拟合未知非参函数,并结合可逆跳MCMC算法、随机游动Metropolis抽样器以及Gibbs抽样技术来对所有参数进行抽样,然后利用数值模拟方法和应用实例考察参数估计精度、未知函数拟合效果以及实际应用效果.结果发现,在两种不同空间数据结构以及多种不同样本量设置下,三个不同分位点上参数估计值的精度较高,非参部分未知函数的拟合效果良好,应用实例也展示了理论方法的实际应用价值.本文的研究结果证实了所提出的模型及其理论方法可以为同时存在线性关系与非线性关系,以及具有厚尾和空间相依特征的变量和数据提供有力的分析工具.This paper proposes a semi-parametric spatial lag quantile regression model,which can simultaneously examine the spatial correlation of the dependent variables and the partial nonlinearity of the influencing mechanism,and can model the response function at any quantile.Secondly,the paper constructs a Bayesian estimation method for the model.In the construction of the Bayesian theoretical framework,the paper uses polynomial splines to fit the unknown nonparametric functions,and samples all parameters by combining reversible jump markov chain monte carlo(RJMCMC) algorithm,random walk Metropolis sampler and Gibbs sampling technique.Then the accuracy of parameter estimation,the fitting effect of unknown function and the effect of practical application are investigated by numerical simulation method and application example.The results show that the accuracy of parameter estimates at three different quantile is higher under two different spatial data structures and a variety of different sample sizes.And the fitting effect of non-parametric unknown functions is good.The practical application of the theoretical method is also demonstrated.The results of this paper demonstrate that the proposed model and its theoretical approach can provide a powerful analytical tool for variables and data with both linear and nonlinear relationships,and with thick tails and spatial dependencies.

关 键 词：半参数 空间滞后模型 分位数回归 贝叶斯估计 

分 类 号：F224.0[经济管理—国民经济]