Bayesian spatial modelling of geostatistical data using INLA and SPDE methods: A case study predicting malaria risk...

Area of Science:

• Environmental Science
• Public Health
• Statistical Modeling

Background:

• Bayesian spatial models are crucial in health, ecology, and environmental science.
• Traditional Markov chain Monte Carlo (MCMC) methods are computationally intensive and problematic for big data.
• Integrated Nested Laplace Approximation (INLA) provides a less intensive alternative for Bayesian inference.

Purpose of the Study:

• To demonstrate fitting Bayesian spatial models using INLA and SPDE approaches.
• To analyze geostatistical data for spatial process prediction and covariate effect assessment.
• To apply these methods to malaria prevalence data in Mozambique using R-INLA.

Main Methods:

• Utilizing the Integrated Nested Laplace Approximation (INLA) for approximate Bayesian inference.
• Employing the stochastic partial differential equation (SPDE) approach for geostatistical data analysis.
• Applying the R-INLA package for fitting and interpreting Bayesian spatial models.

Main Results:

• Successfully fitted a Bayesian spatial model to malaria prevalence data.
• Demonstrated the prediction of malaria risk and assessment of covariate effects.
• Provided R code for reproducibility and application in other spatial analyses.

Conclusions:

• INLA and SPDE offer an efficient computational approach for Bayesian spatial modeling.
• These methods are effective for analyzing geostatistical data and understanding spatial processes.
• The R-INLA package facilitates practical application in public health and environmental research.