Spatio-temporal variation in childhood growth in Nigeria: a comparison of aggregation and interpolation

ABSTRACT

The mean height-for-age Z-score (HAZ) of children under five is an important indicator of the health status of a population. HAZ values are frequently aggregated and reported at the national level, potentially obscuring important within-country variation. We evaluated aggregation and interpolation methods to provide sub-national estimates over space and time, using survey data from Nigeria in 1990, 2003, 2008, and 2013. We aggregated HAZ values by region and by state, and interpolated the values spatially and spatio-temporally using thin plate splines. The results were evaluated with cross-validation using the root mean squared error (RMSE) as a measure of goodness of fit. Mean HAZ values increased from 1990 to 2013, but values rose more sharply in southern Nigeria than in the North. All methods performed better than assuming a constant national average. The state-level aggregation, and the spatial and spatio-temporal interpolations had similar RMSE values, but the interpolation methods showed more detailed spatial variation. Spatio-temporal interpolation produced good results in all conditions, including in years with sparse sampling and when extrapolating to years for which there were no observations.