TMB workshop

Presentation to accompany the workshop can be found here: https://docs.google.com/presentation/d/1foFjojH5EN8MyzqNVSeTafMI8m1BKgfrSer0Bku9qwE/edit?usp=sharing

Helpful links for reference:

Comprehensive TMB documentation: http://kaskr.github.io/adcomp/_book/Introduction.html
Specific help for coming to C++ from R: http://kaskr.github.io/adcomp/_book/CppTutorial.html#i-know-r-but-not-c
Github link with some simple TMB examples: https://github.com/kaskr/adcomp/wiki

Overview

Data used

Malaria prevalence data: Open-access malaria data hosted by the Malaria Atlas Project https://map.ox.ac.uk/ accessed using the malariaAtlas R package

Malaria incidence data: Open-access data from https://datadryad.org/bitstream/handle/10255/dryad.82194/PfPvAllData01042015_AgeStand.csv?sequence=1. Saved in the data folder

Covaraite data: WorldClim data (temperature, temparature variation and precipitation) at 5 minutes resolution from https://www.worldclim.org/bioclim using the getData R function

Structure of the course

Three models of increasing complexity implemented in TMB to predict malaria prevalence.

Model 1: Linear regression with covariates

Model 2: Linear model with Gaussian field

Model 3: Two types of response data. Two contributions to the likelihood. Customised link function. Linear model with Gaussian field and optimised link function

File structure

Three main scripts for 3 different models:

1_example_linear_regression.R
2_example_with_field.R
3_example_with_link_function.R

model specification cpp files (in src folder):

src/model1.cpp
src/model2.cpp
src/model3.cpp

Data preparation:

prepare_data.R: prepares prevalence survey data and covariate data for all 3 models
prepare_incidence_data.R: prepared incidence survey data for 3rd model

Predictions:

predict.R: predict prevalence for model 2 and 3. Plot mean and uncertainty

Needed libraries

If you are coming to the workshop please install these packages ahead of time.

malariaAtlas
dplyr
sp
raster
TMB
INLA
Matrix
ggplot2
cowplot