Details
Date: Mar 12, 2025
Location: San Francisco, CA - USA
Abstract
Accurate modeling of spatial dependence is crucial for analyzing spatial data and influencing parameter estimation and predictions. Existing models for areal data often struggle to differentiate between different polygon shapes, while data fusion models face computational challenges with larger datasets. To address these limitations, we propose the Hausdorff-Gaussian process (HGP), a versatile model using the Hausdorff distance to capture spatial dependence in point and areal data. We introduce a valid correlation function, accommodating various modeling techniques, including geostatistical and areal models, and integrate it into generalized linear mixed-effects models for data fusion. We demonstrate the HGP’s competitive performance regarding goodness-of-fit and prediction through simulations and real-world applications involving both areal data and data fusion. The HGP offers a flexible and robust solution for modeling diverse spatial data with potential public health and climate science applications.