作　　者： ; ; ;

机构地区： 华南理工大学经济与贸易学院

出　　处： 《统计与信息论坛》 2011年第4期14-20,共7页

摘　　要： 基于OLS估计残差,将Bootstrap方法用于空间滞后相关LM-Lag检验。在不同的误差结构和空间权重矩阵条件下,比较Bootstrap LM-Lag检验和渐近检验的水平扭曲和功效。通过Monte Carlo实验表明,当误差项不服从经典正态分布假设时,LM-Lag渐近检验存在严重的水平扭曲,Bootstrap检验能够有效地校正水平扭曲,并且Bootstrap LM-Lag检验的功效与渐近检验近似;无论误差项是否服从正态分布,从水平扭曲和功效角度看,线性回归模型Bootstrap LM-Lag检验有效。 In this paper,the authors applies the Bootstrap methods for LM-Lag to test the spatial lag correlation based on the OLS residuals.The size distortion and power of Bootstrap and asymptotic tests are evaluated and compared for various structures of error and spatial weight matrix.For more realistic heterogeneous non-normal distributional models,the asymptotic tests for LM-Lag statistics perform poorly with large size distortion.Instead,Bootstrap tests have shown superiority in smaller size distortion.In addition,the power of Bootstrap testing is approximately equivalent to that of the asymptotic tests.Furthermore,our extensive Monte-Carlo simulation indicates that in view of size distortion and power,Bootstrap LM-Lag testing for linear regression models is an effective approach whether the classical distributional assumption is violated or not.

关 键 词： 水平扭曲 功效 方法 检验

领　　域： [理学] [理学]