空间断点回归交互效应模型设定方法及其应用

Setting Method of Interactive Effect Model in Spatial Discontinuity Regression and Its Application

空间断点回归实现因果关系统计推断的实验场景决定了交互效应是其固有特征。已有交互效应模型设定方法未充分考虑变量间的空间关系,使其应用局限于微观领域。文章在空间断点回归分析框架下,提出了模型化交互效应的更一般化方法,将其应用推广至区际差异分析等宏观领域,充分讨论了能够识别交互效应的主要统计假设、模型设定方法和统计推断方法,以及忽略交互效应产生的影响。基于此方法,对重庆市升级为直辖市的经济效应进行实验设计、统计推断和评价。实证分析表明,当恰当利用交互效应模型进行统计推断时,处理效应和交互效应估计值均显著为正,处理效应t检验显著性更强,且其估计值在一定空间范围内显著大于普通空间断点回归模型的估计值,该结论基本稳健。

The experimental scenario using spatial discontinuity regression to realize statistical inference of causality deter-mines that the interactive effect is its intrinsic feature.The existing methods for setting up interactive effect models have not fully considered the spatial relationship between variables,limiting their application to micro-level analysis.In the framework of spatial discontinuity regression analysis,this paper proposes a more generalized approach to modeling interactive effects,expands its ap-plication to macro-level analysis such as inter-regional differences,and fully explores the main statistical assumptions,model set-ting methods,and statistical inference methods for identifying interactive effects,as well as the impact of neglecting interactive ef-fects.Based on this approach,an experimental design,statistical inference,and evaluation are conducted on the economic effects of Chongqing upgraded to a municipality directly under the central government.The empirical analysis is shown as follows:When the interactive effect model is properly used for statistical inference,both the estimated values of the treatment effect and the inter-active effect are significantly positive,and the t-test of the treatment effect is more significant,and its estimated value is signifi-cantly greater than that of the common spatial discontinuity regression model within a certain spatial range.The conclusion is basi-cally robust.