Child Mortality Using Bayesian Semi-Parametric Discrete-Time Survival Model

Tesfaye A. Jimma

The Bayesian Approach offers a viable and rigorous solution, though there is also the added benefit of providing much-needed uncertainty and probability assessments in non-linear and non-Gaussian situations in a valid and rigorous way. Mortality and its various determinants have been traditionally studied in a regression modeling framework. Initial studies mostly used the usual linear regression models which, however, are not appropriate in situations where the mortality information is given by a binary indicator of dead or alive. Binary regression models (logit and probit) are, therefore, logical alternatives. There are, however, problems, with logit and probit models, namely, that they do not take into consideration the information on the survival time. Hence, most studies now utilize survival analysis techniques. Recently, a Bayesian Geo-Additive modeling framework which encompasses most of the known regression models and improves upon their shortcomings. The proposed model is also called Bayesian semi-parametric structured regression model.