Probabilistic Projections of Populations With HIV: A Bayesian Melding Approach

Samuel J. Clark, University of Washington
Jason Thomas, University of Washington

Population projection models are valuable tools for demographers and public policy makers alike. A particular example is the model developed by Heuveline (2003), which captures some of the links between population growth and the spread of HIV/AIDS. This model requires relatively few inputs and can provide projections of HIV prevalence for populations for which reliable data are limited. We reproduce Heuveline’s work, but in a Bayesian context. More specifically, we use Bayesian melding to obtain measures of uncertainty around both the model inputs and outputs in the form of probability distributions. This approach provides useful information to policy makers concerning issues with planning and resource allocation.