Projection-Based Approach to Alleviate Spatial Confounding in Spatial Frailty Models

• Activity Number: 554 - Recent Challenges and Developments in Bayesian Big Data Inference and Computation with Public Database
• Type: Invited
• Date/Time: Thursday, August 6, 2020 : 3:00 PM to 4:50 PM
• Sponsor: ENAR
• Author(s): Marcos Oliveira Prates* and Douglas Azevedo and Dipankar Bandyopadhyay
• Companies: Universidade Federal de Minas Gerais and Universidade Federal de Minas Gerais and Virginia Commonwealth University
• Keywords: Spatial Statistics; Spatial Confounding; Survival Analysis; INLA; Spatial Survival Confouding; Restricted Models

Abstract

Spatial Confounding is the name given to the confounding between fixed and spatial random effects in Generalized Linear Mixed Models. It has been widely studied and it gained attention in the past years in spatial literature, as it may generate unexpected results in modeling. The projection-based (restricted models) approach appears like a good way to work around the Spatial Confounding in this family of models. However, when the support of fixed effects is different from the spatial effect, this approach can no longer be directly applied. Spatial Frailty models are able to incorporate spatially structured effects in survival models, and it is common to observe more than one sample unit per area which means that the support of fixed and spatial effects differs. In this work, we introduce a projection-based approach for Spatial Frailty models where the support of fixed and spatial effects do not match. To provide a fast inference for the parameters we used the Integrated Nested Laplace Approximation (INLA) methodology. The Restricted Spatial Frailty model is applied to modeling cases of Lung and Bronchus cancer in California state and the results prove the methodology efficiency.