R codes for INLA-SPDE bottom-up population modelling (Cameroon Application)
Here, we provide the descriptions of the resources within this repository. It contains both the R scripts and all the relevant datasets, except the very large prediction grids covariates (prediction_data.RDS) which is available on Google Drive here: https://drive.google.com/file/d/1290hqUnBHhQS0I3iijddj34solTG-S_S/view?usp=sharing
Additionally, a copy of the Supplementary materials document "Bayesian_Geostat_Pop_Mod_MethodsX_supplementary.pdf" is included.
There are two key scripts contained within the repository to help users reproduce the datasets and to facilitate the implementation of the methodology in other contexts:
-
efficient_pop_model_application_updated.R - this is the main R script developed for the implementation of the methodology using real datasets based on five (5) nationally representative household listing datasets from Cameroon. It contains all the subscripts for initial data exploration, covariates selection through stepwise regression, Bayesian hierachical geostatistical model implementation, model selection, posterior simulations and grid cell predictions, uncertainty quantification, zonal statistics calculation, model cross-validation, and several plots.
-
efficient_pop_sim_study_final.R - this contains the R codes used for the simulation study conducted in the study. The simulation study explored the sensitivity of the methodology to changes in magnitudes of spatial autocorrelation and different proportions of missing data.
-
high_resolution_modelled_estimates_pop_and_hholds.R - this outlines the combined codes used to produce the small area estimates of household numbers and population counts (only used for the combined population count and household count data models).
The input datasets used here are saved as .RData and it contains multiple datasets which are described in the respective scripts.
Please kindly email the lead author on [email protected] or [email protected], should have any feedback or questions. This will be greatly appreciated. Thanks!