CMStatistics 2022: Start Registration
View Submission - CMStatistics
Title: On robustness for spatio-temporal data Authors:  Alfonso Garcia-Perez - Universidad Nacional de Educación a Distancia (UNED). Department of Statistics (Spain) [presenting]
Abstract: The spatio-temporal variogram is an important factor in spatio-temporal analysis because it is the key element in kriging prediction. However, the traditional spatio-temporal variogram estimator, which is commonly used for this purpose, is extremely sensitive to outliers. We address this problem in two different ways. First, defining new robust spatio-temporal variogram estimators, which are defined as M-estimators or trimmed estimators of an original data transformation, estimators for which we obtain accurate approximations for their distributions. Second, we compare the classical estimate against a robust one, identifying spatio-temporal outliers in this way. In these two approaches, we use a multivariate scale-contaminated normal model framework. In the contribution, we also define and study a new class of M-estimators and include real-world applications. We finally determine whether there are significant differences in the spatio-temporal variogram between two temporal lags reducing, if so, the number of lags considered in the spatio-temporal analysis.