Ezra Gayawan (Redeemer’s University): Spatial Bayesian semi-parametric analyses of childhood mortality in Nigeria: estimate from mortality index

We investigate childhood mortality in Nigeria using retrospective data from Demographic and Health Surveys. In order to control for the duration of exposure to the risk of mortality, we compute mortality index; the ratio of child death to expected child death, for each woman based on the ages of her children using United Nations Model Life Table for developing countries. Based on the mortality index, a flexible Bayesian weighted geoadditive regression model that allows for nonlinear, fixed and spatial risk factors was adopted. The nonlinear effects were modelled using Bayesian P-splines while conditional autoregressive prior was assumed for the spatial structure. We assumed diffuse prior for the fixed effects. Inference is fully Bayesian and uses the Markov chain Monte Carlo (MCMC) simulation technique. Results reveal a north-south divide in child mortality in Nigeria and a declining pattern of child mortality with age of mother. The results also provide some evidence on how child mortality can be reduced, by improving socioeconomic and public health conditions and also the need to adopt and implement policies at the district level.
Joint work with Cassio M. Turra