Are continuous models better than areal models for spatial disease mapping with a large number of small areas?
Abstract
Classical methods for analysing areal mortality data typically rely on Poisson mixed models, which account for spatial and unstructured variability through prior distributions such as the BYM2 prior. This paper investigates alternative modeling approaches for settings involving a large number of small areas. Specifically, we explore the Stochastic Partial Differential Equation (SPDE) approach for continuous risk surfaces, examining an existing model from the literature alongside a proposed modification that incorporates an unstructured random effect. Finally, we evaluate the performance of both hierarchical mixed Poisson models and SPDE-based frameworks through a comprehensive simulation study across various spatial configurations and aggregation levels. Furthermore, we apply these approaches to the analysis of lung cancer mortality among men and women across Spanish municipalities during the period 2009--2022.
Related
- Amortized Bayesian Disease Mapping and Boundary Detection on Heterogeneous Spatial Graphs
- Bounds on the maximum number of limit cycles of piecewise linear Lienard systems I. The continuous case
- Continuous logistic Gaussian random measure fields for spatial distributional modelling
- Relative Wasserstein Spatial Depth for Cluster Number Selection in Distributional Data
- Continuous mixtures of Gaussian processes as models for spatial extremes
BibTeX
@misc{salcedo2026continuous,
title = {{Are continuous models better than areal models for spatial disease mapping with a large number of small areas?}},
author = {Ismael Salcedo and Camelia Trandafir and Tomás Goicoa and Sara Martino and María Dolores Ugarte},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.20943},
eprint = {2610.20943},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.20943},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - Are continuous models better than areal models for spatial disease mapping with a large number of small areas?
AU - Ismael Salcedo
AU - Camelia Trandafir
AU - Tomás Goicoa
AU - Sara Martino
AU - María Dolores Ugarte
PY - 2026
DA - 2026/10/14
PB - arXiv
DO - 10.48550/arxiv.2610.20943
UR - https://arxiv.org/abs/2610.20943
AB - Classical methods for analysing areal mortality data typically rely on Poisson mixed models, which account for spatial and unstructured variability through prior distributions such as the BYM2 prior. This paper investigates alternative modeling approaches for settings involving a large number of small areas. Specifically, we explore the Stochastic Partial Differential Equation (SPDE) approach for continuous risk surfaces, examining an existing model from the literature alongside a proposed modification that incorporates an unstructured random effect. Finally, we evaluate the performance of both hierarchical mixed Poisson models and SPDE-based frameworks through a comprehensive simulation study across various spatial configurations and aggregation levels. Furthermore, we apply these approaches to the analysis of lung cancer mortality among men and women across Spanish municipalities during the period 2009--2022.
ER -