CMStatistics 2021: Start Registration
View Submission - CFE
Title: Fitting approaches of Cliff-Ord models to data affected by locational errors Authors:  Flavio Santi - University of Verona (Italy) [presenting]
Diego Giuliani - University of Trento (Italy)
Giuseppe Espa - University of Trento (Italy)
Maria Michela Dickson - University of Trento (Italy)
Abstract: When a spatial regression model is fitted to micro-geographic data in order to account for spatial dependence, the quality of information on unit locations becomes relevant. Locational errors may originate from flaws in geocoding or georeferencing processes, as well as from geomasking of unit positions; as a result, the actual positions of units are known up to an error, whose probabilistic behaviour may be either known or unknown, and whose magnitude may be either homogeneous or heterogeneous amongst units. Spatial regression models a la Cliff-Ord are known to suffer heavily from the consequences of locational errors, as they typically make the parameter estimators markedly biased and inconsistent. A review is made for fitting approaches of Cliff-Ord models to data affected by locational errors. A new hybrid approach is proposed which combines analytical and computational methods.