线性回归模型Bootstrap LM-Error检验的水平扭曲

基于OLS估计残差,本文将Bootstrap方法用于空间误差相关性LM-Error检验,综合考虑Bootstrap模拟抽样次数、空间衔接结构以及样本量,研究并比较空间误差相关Bootstrap LM-Error检验与渐近检验的水平扭曲。大量Monte Carlo实验结果显示,当模型误差不满足独立正态分布的假设条件时,空间误差相关LM-Error渐近检验的水平扭曲较大,采用Bootstrap方法可以较好地降低该水平扭曲;不管模型误差是否满足独立正态分布的假设条件,Bootstrap方法均能够有效地降低LMError渐近检验的水平扭曲。

Size Distortion of Bootstrap LM-Error Tests in Linear Regression Models

In this paper bootstrap methods for LM-Error testing of spatial error autocorrelation based on the OLS residuals axe applied.The size distortion of bootstrap and asymptotic tests axe explored for various spatial structures,numbers of bootstrap and different sample size Our extensive Monte-Carlo simulation indicates that bootstrap tests have the capability of improving size distortion,which exists in the asymptotic tests for LM-Error statistics under considering the more realistic non-normal distribution assumption;whether the errors in the model have the independent normal distribution or not,bootstrap methods can effectively reduce the size distortion of LM-Error asymptotic test.