变系数空间自回归模型的Bootstrap检验

A Bootstrap Test for the Varying Coefficient Spatial Autoregressive Models

变系数空间自回归模型是变系数模型在空间数据分析方面的推广,因其众多的应用背景而得到广泛的重视和研究,确认模型中系数是否真正随变量的变化而变化是应用变系数空间自回归模型需解决的首要问题.本文基于Bootstrap检验方法研究了变系数空间自回归模型中的常系数项的辨别问题,为建立半变系数空间自回归模型提供依据.最后,通过模拟试验验证Bootstrap检验方法在有限样本容量下的有效性.数值模拟分别考察了误差项服从不同分布、空间滞后相关系数变化以及解释变量共线性程度不同时,Bootstrap方法逼近其零分布的准确性以及检验的功效.模拟结果表明本文所提出的Bootstrap方法能精确地逼近检验统计量的零分布且检验具有满意的功效.

The varying coefficient spatial autoregressive model is a generalisation of a varying coefficient model in spatial data analysis.It has been widely valued and studied because of its many application backgrounds.To apply the models,the first problem is to confirm whether the coefficients in the model change with respect to a variable.Based on the Bootstrap test,the identification of constant-coefficient terms in the varying coefficient spatial autoregressive models is studied,which provides a basis for the confirmation of a semi-varying coefficient spatial autoregressive model.Furthermore,some simulation experiments are conducted to evaluate the validity of the Bootstrap approximation in the case of a finite sample size.Meanwhile,when the error term distribution is different.The value of the spatial autoregressive parameter changes,and the collinearity among the explanatory variables varies,the accuracy of the bootstrap approximation to their null distributions and the power of the test are investigated.The simulation results demonstrate that the proposed Bootstrap method can accurately approximate the zero distribution of test statistics,and the test has a good effect.