Cross-validation results showing the mean of the distribution of mean absolute error (MAE) for three mortality indicators and amongst all-ages and life tables after 1,000 iterations of cross validation.

HIV MLT refers to the model presented in this paper with three possible input combinations: 1) HIV prevalence with life expectancy at birth [input: ], 2) HIV prevalence with child mortality [input: ], or 3) HIV prevalence with child mortality and adult mortality [input: and ]. MAE for ‘All-ages’, , and expressed per 1,000 and the smallest number in each column is bolded. For HIV MLT model, ‘–’ indicates MAE 0.001 (or 0.01 for ).

‘All-ages’ refers to the mean absolute error for the non-logged mortality rates across age groups (0, 1–4, 5–9, 10–14,…, 75) and amongst all life tables ().

‘WHO’, ‘CD’, and ‘Log-Quad’ contain blank spaces as these quantities are inputs to these systems and thus have no error.