English  Français

Trends in chronic child undernutrition in Bangladesh for small domains using Bayesian hierarchical time series modelling

Sumonkanti Das, Australian National University
Bernard Baffour, Australian National University
Alice Richardson, The Australian National University ACT 2601

Chronic undernutrition in children is an important population health problem with adverse outcomes in both the short and longer term. In order to provide an evidence base for the trend towards these goals in Bangladesh, this study aims to estimate trends of child chronic undernutrition (stunting) in that country for small domains cross-classified by five children age-group (0-11, 12-23, 24-35, 36-47, 48-59 months) and sixty-four districts. Bayesian multilevel time-series models were developed based on micro-data extracted from seven Bangladesh Demographic and Health Surveys conducted in from 1996 to 2017. Direct estimates of stunting and their standard errors for the target small domains were calculated from the data then used as input for the time series models. These models borrow cross-sectional, temporal, and spatial strength in such a way that they can interpolate stunting levels in non-survey years for all small domains. Estimates for higher aggregation levels are obtained by aggregation of the small domain predictions. Results show that the national-level trend is in steep decline over the period, from about 3-out-every-5 to about 1-in-3 children. At a small domain level (age x geographical area x time) we find more nuanced differences, and improvements through using Bayesian multilevel time series approaches.

Keywords: Multi-level modeling, Small area estimation, Bayesian methods / estimation, Spatial dependence/heterogeneity

See extended abstract.

  Presented in Session P11.