Application of a Singular Value Decomposition-Based Factorization and Parsimonious Component Model of Mortality to HIV Epidemics in Africa

Jonathan Muir , The Ohio State University
Samuel J. Clark, The Ohio State University

Many countries around the world, particularly in Africa, require the modeling of mortality in order to derive complete age schedules of mortality. These age schedules are required for population estimation, forecasting, and projection and other tasks in Demography and Epidemiology. One method for constructing these models that is garnering increasing interest is the use of singular value decomposition-base factorization (SVD). Using simulated demographic projection data of HIV epidemics calibrated and organized in SPECTRUM, this study applies SVD to (1) derive a parsimonious component model of demographic age schedules using age-specific mortality and (2) predict age-specific mortality using HIV indicators and summary measures of age-specific mortality. The component model of age-specific mortality successfully reproduces the data with four inputs, and acting through those four inputs, HIV related covariates are able to accurately predict full age schedules.

See extended abstract

 Presented in Session 5. Health & Mortality 1