Bayesian modeling of space and time dynamics: A practical demonstration in social and health science research

Objective: This article introduces Bayesian spatial–temporal modeling for social and health science research. We use the World Bank’s World Development Indicators data on youth unemployment and HIV risk in Africa to illustrate the utility of the Bayesian paradigm in modeling space–time changes in outcomes. Method: Data on adolescents and young adults were collected in 36 African countries from 1991 to 2014. We examined associations between HIV risk and youth unemployment rates using 16 Bayesian Poisson models incorporating spatial and temporal autocorrelations. Results: The best fit to the data was the model with spatially uncorrelated heterogeneity, temporally correlated random-walk autocorrelation, and spatial–temporal interaction. HIV risk factors are spatially uncorrelated across 36 countries but temporally correlated (i.e., country and time interaction) over the data collection period. The relationship between HIV risk and unemployment rate is statistically nonsignificant because of large spatial–temporal variations. Conclusions: This article demonstrates the capacity of Bayesian modeling to incorporate spatial (neighborhood) and temporal (historical) information to reflect not only the influences of space and time but also their interactions on the phenomenon of interest. The Bayesian framework holds great promise for improving the dynamic targeting of interventions and strategies to achieve desired outcomes.