Outlier Robust Estimation In Spatial Autoregressive Models

Spatial autoregressive(SAR)model has been extensively used in spatial data modeling.However,outliers are usually encountered in when analyzing spatial data.Outliers tend to impact the parameter estimation and corresponding statistical inference procedure.Therefore,it is necessary to consider more appropriate and stable estimation method under current setting.Robust estimation produces estimators that are robust to the presence of potential outliers.The current study aims to establish the corresponding robust estimation procedure to accommodate the effect of outlying observations.In current thesis,we adopt Huber's rho function to establish the robust maximum likelihood estimation procedure within the framework of maximum likelihood estimation.The corresponding estimation equations are provided accordingly.The Fisher scoring iteration algorithm is also established.Simulations are conducted to assess the performance of the proposed method in finite sample setting.Extensive simulations show that when the data are contaminated by outliers,the proposed robust estimation method leads to better estimation accuracy than conventional maximum estimation method.When there are no outliers,the performances of the two methods are very comparable..We use Anserlin's data to demonstrate the effectiveness of the proposed method and the performance of the robust estimation procedure is also promising in general.