空间混合自回归模型的局部影响分析

Local influence assessment for the spatial mixed autoregressive model

基于一阶导数方法,对空间混合自回归模型进行局部影响分析.当模型中误差向量的均值发生扰动时,依据极大似然方法对模型中自回归系数ρ和方差σ2分别构造了检测强影响点或异常点的最大影响方向dmax,ρ和dmax,σ2.数据模拟研究表明,基于dmax,σ2的检测效果明显优于dmax,ρ的效果.同时,对一个实际数据的分析,说明所得结果在实际研究中也是有用的.

The paper focuses on the study of the local influence assessment for the spatial mixed autoregressive model based on the first -order derivative method. When the mean of the error vector in the model is perturbed,the related maximum influence directions dmax, p and dmax,σ^2 for detecting the influential observations or outliers are established for the autoregressive parameter p and variance σ^22 of errors, respectively. The simulation study shows that the assessment based on dmax,σ^2 is more effective than that on dmax,ρ, An illustrative example is proposed to demonstrate its validity.