Examples of probable measurement error: especially months of age five and 11 (Niger, upper panel) and 11, 13 and 15 (Ghana, lower panel). Data markers show oldest month of age for the observation. High variability could be due to limitations of 24-hour recall, but consistently high age-specific values suggest systematic measurement error.
Lauer et al. BMC Medicine 2004 2:26 doi:10.1186/1741-7015-2-26