Title: Spatial multiscale analysis of functional count models
Authors: Antoni Torres - University of Malaga (Spain) [presenting]
Maria Pilar Frias Bustamante - University of Granada (Spain)
Jorge Mateau - Universi ty Jaume I of Castellon (Spain)
Maria Dolores Ruiz-Medina - University of Granada (Spain)
Abstract: A heterogeneity analysis is performed through different scales to describe the local variability of the sample paths of an infinite-dimensional spatial intensity in the context of a log-Gaussian Cox count model. A more flexible Functional Data Analysis (FDA) preprocessing procedure is implemented, allowing a higher degree of local singularity, avoiding over-smoothing. The derived multiresolution approximation reflects the space-time interaction at different scales, affecting the evolution of the log-intensity process that governs the counts. A wavelet-based nonparametric framework is adopted for spatial functional prediction. The approach presented is implemented for spatiotemporal prediction of respiratory disease mortality at Spanish provinces.