|
Abstract:
|
The analysis of spatial data mainly focuses on the covariance structure and the parameters involved, which are typically unknown and need to be estimated by maximum likelihood estimation. However, the presence of outlying observations in the data might significantly affect the maximum likelihood estimators of the covariance parameters and the associated uncertainties; consequently, leading to inaccurate predictions. In this research project, we are developing a robust method that does no harm to the data and gives reasonable estimates in the presence of outliers. Through an intensive numerical study, we observe that the estimates obtained from the robust model are close to the true values even in the presence of outliers.
|
