摘要
本文对半参数变系数回归模型,构造了新的空间相关性检验统计量,利用三阶矩χ2逼近方法导出了其检验p-值的近似计算公式,蒙特卡罗模拟结果表明,该统计量在检测空间相关性方面具有较高的准确性和可靠性。同时考察了误差项服从不同分布时的检验功效,体现出该检验方法的稳健性。进一步,我们还给出了检验统计量的Bootstrap方法以及检验水平的模拟效果。
Based on the semi-parametric varying coefficient regression model,a new testing statistic for spatial correlation is constructed. The corresponding approximation of p-value is derived by employing the third-moment χ2method.The Monte Carlo simulation studies show that the proposed testing statistic has good accuracy and reliability. Meanwhile,we investigate the simulation results of testing powers for different error terms,which show that the testing method has high robustness. Furthermore,the bootstrap method and simulation results for the testing statistic are given.
出处
《统计研究》
CSSCI
北大核心
2015年第7期87-92,共6页
Statistical Research
基金
教育部人文社会科学项目“空间自回归单指数模型的理论和实践”(13YJA9100002)的阶段性成果
