空间滞后模型的标准化稳健检验及其功效对比分析

Standardized robust LM test for spatial lag model and efficiency analyses

在误差项分布未知、模型设定偏误以及不同形式空间布局的复杂条件下,本文提出检验空间滞后模型的标准化稳健LM检验方法,并采用Bootstrap法和Monte Carlo法对统计量的渐进性进行分析.结果表明:采用Bootstrap法和Monte Carlo法的检验功效均优于现有LM检验方法;对于小样本,结合Bootstrap法与本文提出的统计量进行模型设定的检验更可靠;在空间布局为Group规则时,采用本文提出的统计量进行模型设定的检验效果更好.

This paper proposes a new robust LM test for spatial lag model that is set when there are un- known error distribution, model misspecification and changes of spatial layout. We analyze the asymptotic properties using Bootstrap and Monte Carlo methods which are found to be superior to the LM method. Moreover, we also find that Bootstrap is most efficacious. It is more reliable to use combined Bootstrap and our proposed testing statistic when sample size is small. It is also more reliable to use our proposed testing statistic when ＂Group＂ in spatial layouts is contiguous.