空间自回归模型的变量选择  

Variable Selection for Spatial Autoregressive Models

作　　者：郭双[1] 魏传华[1] 

机构地区：[1]中央民族大学理学院,北京100081

出　　处：《中央民族大学学报（自然科学版）》2015年第3期92-96,共5页Journal of Minzu University of China(Natural Sciences Edition)

基　　金：国家自然科学基金项目(No.11301565);北京高等学校"青年英才计划"项目(No.YETP1316)

摘　　要：空间自回归模型是空间计量经济学中用以刻画观测单元之间空间相关性的常用方法.基于Adaptive Lasso方法研究了这类模型的变量选择问题.利用数值模拟考察了所提方法在有限样本下的表现,结果表明所用的方法可以有效地筛选出不显著的变量,并且给出对因变量影响显著的自变量对应的系数估计值.最后利用该方法分析了波士顿房价数据.Spatial autoregressive models have been commonly used in spatial econometrics to characterize the spatial correlation among spatial units. This paper focused on the variable selection of spatial autoregressive models via Adaptive Lasso method. A simulation study was carried out to assess the performance of the proposed variable selection procedure in finite sample. The results showed that the method could effectively filter out insignificant variables, and give the corresponding coefficient estimates of significant variables. Finally, Boston housing price data were analyzed using this method.

关 键 词：空间自回归模型 变量选择 自适应Lasso 

分 类 号：O212.1[理学—概率论与数理统计]