B0620
Title: Linking climate and dengue using the integrated nested Laplace approximation and the SPDE approach
Authors: Stephen Jun Villejo - University of the Philippines (Philippines) [presenting]
Abstract: The primary goal is to perform a combined analysis of two spatially misaligned data using the stochastic partial differential equations approach (SPDE) estimated using the Integrated Nested Laplace Approximation (INLA) method. In particular, the two datasets considered are measurements of several climate indicators from several weather stations and the incidence of dengue by province in the Philippines. The former data is point-level while the latter is area-level. A continuously-indexed Gaussian process for the climate variables is assumed and is approximated by a discretely-indexed Gaussian Field (GF) via triangulation with the additional Gaussian Markov Random Field assumption. The obtained estimate of the GF is projected on blocks, which are subsets of the 2D space, corresponding to the provincial boundaries in the entire domain area. The Besag-York-Mollie (BYM) specification is used in providing a link between the incidence of dengue and the projections of the Gaussian process for all the climate variables. The results show a significant association between relative humidity and temperature with the incidence of dengue. Moreover, maps of the relative risks and excess risks were also computed using the INLA and Markov Chain Monte Carlo (MCMC) methods with the BYM model as the baseline model. The MCMC method had some convergence issues, but the estimates of the relative risks and excess risks from the two methods are almost the same.
