Food insecurity remains one of the greatest challenges in Africa, hindering the economic development. Related indicators have been almost invariably monitored in the achievement of the Millennium Development Goals. Women’s body mass index (BMI) measures women's nutritional status, a key indicator of the socio-economic development of a country. Despite recent intervention programmes, geographic and socio-economic disparities remain in the BMI distribution. Therefore, it would be important to rely on accurate estimates of women’s mean BMI levels across domains. We consider a small area model with area-specific random effects that capture the large regional differences in BMI levels. We propose a Bayesian model to investigate the role on BMI of a number of socio-economic characteristics such as age, wealth, parity, education, while accounting for regional variation. Since it is reasonable to assume that some of these variables are measured with error, we develop a suitable methodology and investigate the effect of neglecting measurement error in covariates on the assessment of the regression effects and on the prediction of area-specific BMI mean levels.

We apply the proposed model to 2011 DHS data to explore the geographic variability of the BMI in different regions and compare the determinants of women’s nutritional status.