A BAYESIAN SIMULATION MODELING APPROACH TO PREDICTING MATERNAL AGE-SPECIFIC INFANT SURVIVAL OUTCOMES WITH INCOMPLETE DATA

Abstract

Predicting infant survival rates using the Classical Binary Logistic Regression Model with maternal and child characteristics as covariates can be a challenge when the modeler requires a Maternal Age-Specific model but that is not forthcoming. Reason being that mother age is not a significant covariate in the model. One way out of this is to group mother age at child birth into class intervals of age groups and see whether infant survival outcomes vary significantly with the age groups. If this is true, then Classical Binary Logistic Regression Models one for each age group, can be fitted for predicting infant survival outcomes. A fresh challenge would be that of incomplete data since the data set would have been merged by the age groupings. This new challenge can be overcome by the Bayesian Simulation Modeling Approach. Hence our task in this study is to develop a Bayesian Simulation Modeling Procedure implemented on the Simulation package; Windows Bayesian Inference Using Gibbs Sampling (WINBUG) with the aim of modeling the relationship between Infant Survival Outcomes and Maternal and Child characteristics, for each maternal age group. Besides the successful model fit in the face of incomplete data, the overall result of the study revealed that, the three maternal age groups; 15 – 25 years, 26 – 35 years and 36 years and above have positive impact on infant survival rate, while only the weight of infants delivered by mothers who are 36 years and above pose as risk factor to infant survival rate.