摘要
当误差项不服从独立同分布时,利用Moran's I统计量的渐近检验,无法有效判断空间经济计量滞后模型2SLS估计残差间存在空间关系与否。本文采用两种基于残差的Bootstrap方法,诊断空间经济计量滞后模型残差中的空间相关关系。大量Monte Carlo模拟结果显示,从功效角度看,无论误差项服从独立同分布与否,与渐近检验相比,Bootstrap Moran检验都具有更好的有限样本性质,能够更有效地进行空间相关性检验。尤其是,在样本量较小和空间衔接密度较高的情况下,Bootstrap Moran检验的功效显著大于渐近检验。
The asymptotic distribution of Moran＇s I statistic can＇t effectively test spatial correlation among 2SLS residuals in spatial econometric autoregressive models with the i.i.d.error.In this paper,we apply two residual-based Bootstrap methods for diagnostic testing spatial correlation in a spatial econometric autoregressive model.In comparison with the theoretical asymptotic test,our extensive Monte Carlo simulation indicates that in view of power whether the errors are i.i.d or not,bootstrap test for this model has superior finite sample properties,and can more effectively check spatial dependence than the asymptotic test.Especially,the power of Bootstrap test is more remarkable than that of asymptotic test in cases of small sample and Queen spatial weight matrix.
出处
《统计研究》
CSSCI
北大核心
2010年第9期91-96,共6页
Statistical Research
基金
国家自然科学基金项目"空间经济计量模型中Bootstrap方法有效性研究"(70871041)阶段性成果
