空间混合效应随机前沿模型估计方法研究  

The Estimation for Spatial Mixed Effect Stochastic Frontier Models

作　　者：蒋青嬗 韩兆洲[2] JIANG Qing-shan;HAN Zhao-zhou(College of Mathematics and Statistics, Guangdong University of Foreign Studies, Guangdong Guangzhou 510006, China;Economic College, Jinan University, Guangdong Guangzhou 510632, China)

机构地区：[1]广东外语外贸大学数学与统计学院,广东广州510006 [2]暨南大学经济学院,广东广州510632

出　　处：《数理统计与管理》2019年第4期628-638,共11页Journal of Applied Statistics and Management

基　　金：全国统计科学研究项目(2018LY81);教育部人文社会科学研究规划基金项目(16YJA910001);广东省自然科学基金(2018A030310572)

摘　　要：在经典随机前沿模型中引入随机效应和空间效应,构建了空间混合效应随机前沿模型。模型考虑了生产单元技术水平的异质性并且可以有效避免忽略内生性问题导致的有偏且不一致估计量,适用性更佳。使用贝叶斯方法估计模型,核心在于推导未知参数的后验分布以及MCMC抽样。相比于其他估计方法,贝叶斯方法简单直观且精度较高,更适合复杂模型的估计。数值模拟结果显示:①贝叶斯估计的精度较高。增加样本容量有助于提高精度。②忽略随机效应,估计精度偏低。数值模拟表明文中模型和方法有存在必要性。This paper constructs the spatial mixed effect stochastic frontier model by incorporating random effects and spatial effects into classical stochastic frontier model.The model takes the heterogeneity of technological level into consideration and can effectively avoid biased and inconsistent estimators derived from endogeneity.Compared with the traditional stochastic frontier model,the model is more adaptable.Use Bayesian method to estimate the model,the core processes are the deduction of posteriori distributions and MCMC sampling.Compared with other estimation methods,the Bayesian estimation is simpler and with higher accuracy which is more suitable for the estimation for complex models.The simulations show that:(i)The accuracy of Bayesian estimation is high.With sample size increasing,the accuracy gets higher,(ii)Neglecting random effects,the accuracy becomes low.The model and estimation method in this article are necessary.

关 键 词：随机前沿模型 空间效应 随机效应 极大似然估计 贝叶斯估计 

分 类 号：O212[理学—概率论与数理统计]