Subnational Under-Five Mortality Rates with Age-Period-Cohort models

Summary

This document provides the code to run the analysis for the manuscript Estimating subnational Under-Five Mortality Rates (U5MR) using a spatio-temporal Age-Period-Cohort model. Within the repository, there are seperate folders for the Code and additional Information on Kenya used within the code.

Data

• Kenyan U5MR data: we cannot share the 2014 Kenyan Demographic and Health Surveys (DHS) data. To access this, the user will need to sign up to the DHS with a suitable project that allows access to the Kenyan dataset specifically.
• Kenyan spatial polygons: these can be downloaded from the Database of Global Administrative Areas (GADM)
• Population totals: the population totals for the proportional aggregation were calculated using the World Population

Code

0. Pre-analysis before running code for analysis

• Run the createDirectoryStructure.R folder to create all the folders where data will be stored
• Download the shapefiles from GADM
• Get access to the Kenyan DHS after registration with the DHS

1. Run dataProcessing.R to download the 2014 KDHS and organise it into the correct format
2. Defining the proportions for aggregation:

• Run adminWeights.R to download the yearly-regional proportions from worldpop
• Find the most recent census for Kenya and write a .csv file with the most recent urban and rural fractions for each region
• Run urProportions.R to assign the correct names to the urban/rural fractions in the .csv file
• Run urThreshold.R to create the yearly-regional-urban/rural proportions using the yearly-regional proportions and the urban/rural fractions as described by Wu and Wakefield

3. Run dataExploration.R to generate plots of the cluster locations and population proportions
4. Generate estimates:

• Run summerEstimates.R to generate and save the direct and Fay-Herriot estimates for Kenyan U5MR
• Run apcEsimates.R to generate and save the Age-Period, Age-Cohort and Age-Period-Cohort subnational model estimates
• Run crossValidation.R to perform the cross-validation for the Age-Period, Age-Cohort, and Age-Period-Cohort models

5. Run comparisonPlots.R to make any plots and tables seen in the manuscript and supplementary material