Estimating Sub-National Female Sex Worker Population Sizes in Sub-Saharan Africa with Regression Analysis
Open Access
Author:
Chen, David
Area of Honors:
Statistics
Degree:
Bachelor of Science
Document Type:
Thesis
Thesis Supervisors:
Le Bao, Thesis Supervisor Le Bao, Thesis Honors Advisor Xiaoyue Niu, Faculty Reader
Keywords:
HIV Linear mixed-effects model Female sex workers
Abstract:
Across sub-Saharan Africa, female sex workers (FSW) suffer from HIV at disproportionally higher rates than the general public. Despite their importance in addressing the HIV epidemic, many countries do not have sufficient data regarding the number of sex workers in their country to inform policy making. By conducting literature review to obtain existing size estimates, and collecting widely available spatial covariates and national predictors, we conducted a linear mixed-effects regression model to predict administrative level one (districts, regions, etc.) FSW size estimates across all of sub-Saharan Africa. Following principal component analysis to address predictor multicollinearity, stepwise selection to perform variable selected, and Leave-One-Country-Out (LOCO) cross-validation to evaluate model performances, a final model consisting of the country, region, and spatial variables was selected. With this model, countries and regions that do not have size estimates can use these model predictions to obtain a preliminary FSW count, helping to inform future funding allocation, outreach, and research efforts.
