A coregionalization model can assist specification of Geographically Weighted Poisson Regression: Application to an ecological study.
Manuel Castro Ribeiro, António Jorge Sousa, Maria João Pereira
Author Information
Manuel Castro Ribeiro: Centro de Recursos Naturais e Ambiente, Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais, 1049-001 Lisboa, Portugal. Electronic address: manuel.ribeiro@tecnico.ulisboa.pt.
António Jorge Sousa: Centro de Recursos Naturais e Ambiente/DECivil, Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais, 1049-001 Lisboa, Portugal. Electronic address: ajsousa@tecnico.ulisboa.pt.
Maria João Pereira: Centro de Recursos Naturais e Ambiente, Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais, 1049-001 Lisboa, Portugal. Electronic address: maria.pereira@tecnico.ulisboa.pt.
The geographical distribution of health outcomes is influenced by socio-economic and environmental factors operating on different spatial scales. Geographical variations in relationships can be revealed with semi-parametric Geographically Weighted Poisson Regression (sGWPR), a model that can combine both geographically varying and geographically constant parameters. To decide whether a parameter should vary geographically, two models are compared: one in which all parameters are allowed to vary geographically and one in which all except the parameter being evaluated are allowed to vary geographically. The model with the lower corrected Akaike Information Criterion (AICc) is selected. Delivering model selection exclusively according to the AICc might hide important details in spatial variations of associations. We propose assisting the decision by using a Linear Model of Coregionalization (LMC). Here we show how LMC can refine sGWPR on ecological associations between socio-economic and environmental variables and low birth weight outcomes in the west-north-central region of Portugal.
