Title: New multivariate spatio-temporal P-spline models for areal count data
Authors: Tomas Goicoa - Universidad Publica de Navarra (Spain) [presenting]
Maria Dolores Ugarte - Universidad Publica de Navarra (Spain)
Gonzalo Vicente - Universidad de Cuyo (Argentina)
Abstract: Univariate spatio-temporal models for estimating risk or rates have been extensively used in disease mapping, mainly to study certain diseases such as cancer. However, and despite their potential, multivariate models are not so widespread in practice due to computational burden and difficulties in implementation. We propose multivariate spatio-temporal P-spline models for areal count data with special emphasis on crimes against women, a public health problem of epidemic proportions according to the World Health Organization. The joint modelling of the different crimes improves the precision of the estimates in comparison to univariate models and provides between-crime correlations. More precisely, correlations between the spatial patterns and the temporal trends of the different crimes are obtained. The models are fitted using Integrated nested Laplace approximations (INLA) and are implemented in R using the generic construction in the package R-INLA. The methodology is used to analyze four crimes against women in the Indian state of Maharashtra during the period 2001-2013.