B1045
Title: Accounting for geomasking in spatial modelling of complex survey data: A data fusion approach
Authors: John Paige - NTNU - Norwegian University of Science and Technology (Norway) [presenting]
Geir-Arne Fuglstad - NTNU - Norwegian University of Science and Technology (Norway)
Andrea Riebler - NTNU - Norwegian University of Science and Technology (Norway)
Abstract: Positional error, error in the locations of spatial data, can bias a spatial model's parameter estimates and spatial predictions when improperly accounted for, and is relevant in applications from public health to paleoseismology. Existing methods that account for positional error frequently either rely on non-generalizable parametric assumptions, employ ad hoc techniques, or use computationally intensive MCMC. A newly introduced method addressing these issues is shown to be extended to account for arbitrary positional error distributions including jittering and geomasking, the censoring of each observation point location up to the area containing it. A flexible numerical integration scheme is further provided, accounting for spatial covariate information. The proposed method has been applied the method to women's secondary education completion data in the 2018 Nigeria demographic and health survey (NDHS) containing point locations jittered via random radial displacements, and the 2016 Nigeria multiple indicator cluster survey (NMICS) containing geomasked locations. Both surveys add positional errors intentionally for confidentiality purposes. In this setting where high-quality survey data is sparse, it is shown via validation that the spatial fusion of these two datasets in a statistically rigorous way improves parameter estimates and spatial prediction precision.
