Analysis of Nigerian Children Nutritional Status Using Bayesian Models

A study into geographical variability of nutritional status of children in Nigeria was observed from geostatistical mapping or kriging and a number of continuous covariates, such as height for age (stunting), weight for height (wasting) and weight for age (underweight) that exhibit pronounced non-linear relationships with the response variable was analysed. To properly account for stunting, wasting and underweight effects on child's age, sex, their place of resident, mothers' educational levels, parents' wealth index, regions and state of the child, kriging and additive models were merged using modified Cox model. The resulting Generalized Additive Mixed Model (GAMM) representation for the geoadditive model allows for fitting and analysis using BayesX software. The Multiple Indicator Cluster Survey 3 (MICS3) data set contains several variables. Only those that are believed to be related to nutritional status were selected. All categorical covariates are effect coded. The child's age is assumed to be nonlinear; the state is spatial effect while other variables are parametric in nature.