Bayesian variable selection in modelling geographical heterogeneity in malaria transmission from sparse data: an application to Nouna Health and Demographic Surveillance System (HDSS) data, Burkina Faso

Background Quantification of malaria heterogeneity is very challenging, partly because of the underlying characteristics of mosquitoes and also because malaria is an environmentally driven disease. Furthermore, in order to assess the spatial and seasonal variability in malaria transmission, vector data need to be collected repeatedly over time (at fixed geographical locations). Measurements collected at locations close to each other and over time tend to be correlated because of common exposures such as environmental or climatic conditions. Non- spatial statistical methods, when applied to analyze such data, may lead to biased estimates. We developed rigorous methods for analyzing sparse and spatially correlated data. We applied Bayesian variable selection to identify the most important predictors as well as the elapsing time between climate suitability and changes in entomological indices.