Fig 1.
Percentage of under-5 deaths at ages 1 to 23 months in DHS surveys.
Percentage estimated for each of the 204 selected surveys. Neonatal deaths not represented in the graph.
Fig 2.
Cumulative probability of dying from birth to age x for the Bolivia 1989 DHS.
Fig 3.
Data versus model predictions in the relationship between infant and child mortality.
U5MD-VR is for Under-5 Mortality Database vital records. For the regional classification of DHS, CSSA is for Central, South, and Southeast Asia, LAC is for Latin America & Caribbean, NA/WA/E is for North Africa/West Asia/Europe, SSA is for Sub-Saharan Africa.
Table 1.
Maximum values of mean error of predictions (%) defining a good fit of the k-model.
Table 2.
Numbers and percentages of surveys included in the analysis according to the 90% and 99% maximum values of fitting error.
Fig 4.
Predicted vs observed ratio of the IMR in individual DHS surveys by world region with predictions made using the k-model or MLTs.
CSSA is for Central, South, and Southeast Asia. LAC is for Latin America & Caribbean. NA/WA/E is for North Africa/West Asia/Europe. Sub-Saharan Africa results are not showed because they include only one survey.
Fig 5.
Predicted vs observed ratio of the IMR in individual DHS surveys by decade with predictions made using the k-model or MLTs.
Fig 6.
Examples of surveys with very good fit of the k-model.
Prediction errors are within the 90% maximum values. The k-model was fitted to the observed q(x) values using ages 0–8 and 24–60 months only. The ticker vertical line indicates the comparison between the observed and predicted infant mortality rate (IMR).
Fig 7.
Examples of surveys with relatively good fit of the k-model.
Prediction errors are between the 90% and 99% maximum values. The k-model was fitted to the observed q(x) values using ages 0–8 and 24–60 months only. The ticker vertical line indicates the comparison between the observed and predicted infant mortality rate (IMR).
Fig 8.
Examples of surveys excluded from analysis due to poor fit of the k-model.
Prediction errors are outside the 99% maximum values. The k-model was fitted to the observed q(x) values using ages 0–8 and 24–60 months only. The ticker vertical line indicates the comparison between the observed and predicted infant mortality rate (IMR).
Fig 9.
Mean prediction error (%) of k-model among excluded surveys by region.
CSSA is for Central, South, and Southeast Asia. SA is for Sub-Saharan Africa.