网络结构数据空间回归模型的平均估计  

Model Averaging Estimation of Spatial Regression Model with Network Structure Data

作　　者：孙志猛[1] 马倩雯 李潇宁 SUN Zhirneng;MA Qianwen;LI Xiaoning(School of Statistic and Mathematics,Central University of Finance and Economics,Beijing 100081;School of Insurance,Central University of Finance and Economics,Beijing 100081)

机构地区：[1]中央财经大学统计与数学学院,北京100081 [2]中央财经大学保险学院,北京100081

出　　处：《系统科学与数学》2018年第6期662-678,共17页Journal of Systems Science and Mathematical Sciences

基　　金：国家自然科学基金(11301561,71473279);国家社会科学基金重大项目(15ZDC024);中央财经大学青年英才培育支持计划项目(QYP1810),中央财经大学社会调研与科研数据库项目,中央财经大学第三批青年科研创新团队项目,中央财经大学学科建设经费资助课题

摘　　要：采用空间误差模型对数据的网络结构关系进行刻画,考虑了空间误差模型的S-AIC和S-BIC模型平均估计方法,证明了S-AIC和S-BIC估计的相合性和渐近正态性.通过蒙特卡洛模拟试验,研究了所提估计的有限样本性质,模拟结果显示,S-AIC和S-BIC模型平均估计表现优于AIC和BIC模型选择估计.利用文章所提方法,对QQ用户数据进行实证分析,说明了所提方法在实际问题中的应用价值.In this paper, we adopt the spatial error model to depict the network structure relationship between individuals, we consider S-AIC and S-BIC model averaging estimation of the spatial error model and show the consistency and asymptotical normality of the S-AIC and S-BIC estimators. We conduct Monte Carlo experiments to investigate the finite sample properties of the proposed estimators. The simulation results show that the S-AIC and S-BIC model averaging estimators perform better than the AIC and BIC model selection estimators. We analyze the QQ user data set to illustrate the application of the proposed method.

关 键 词：空间误差模型 模型选择 模型平均 光滑AIC 光滑BIC 

分 类 号：O212.1[理学—概率论与数理统计]