# A General Age-Specific Mortality Model with An Example Indexed by Child or Child/Adult Mortality

**Authors:** Samuel J. Clark  
**arXiv:** [1612.01408](https://arxiv.org/abs/1612.01408)  
**Date:** 2016-12-01  
**Summary page:** https://www.summarizepaper.com/en/paper/1612.01408v1/

## Key points

- Mortality models are crucial in many countries, especially in Africa
- These models are essential for population estimates, projections/forecasts, and other tasks in demography and epidemiology
- The ability to relate child mortality to mortality at other ages is an important aspect of these models
- The objective of the study was to design a general model for age-specific mortality that provides a standard way to relate covariates to age-specific mortality, calibrate that model using the relationship between child or child/adult mortality and mortality at other ages, and validate the calibrated model and compare its performance to existing models
- The researchers designed a general, parametrizable component model of mortality using singular value decomposition (SVD-Comp) and calibrated it using the relationship between child or child/adult mortality and mortality at other ages observed in the Human Mortality Database
- Cross-validation was used to validate the model's predictive performance compared to that of the Log-Quad model designed for similar purposes
- The results showed that the child-mortality-calibrated SVD-Comp accurately represented observed mortality schedules in the Human Mortality Database and outperformed existing models such as Log-Quad Model
- This approach could be used as a useful tool where child-mortality data is available but information on other age groups' mortalities is unknown.

## Summary

The need for mortality models is crucial in many countries, particularly in Africa where the majority of countries require them to estimate complete age schedules of mortality. These models are essential for population estimates, projections/forecasts, and other tasks in demography and epidemiology. One important aspect of these models is their ability to relate child mortality to mortality at other ages since all countries have measures of child mortality. In this study, the objective was threefold: 1) design a general model for age-specific mortality that provides a standard way to relate covariates to age-specific mortality; 2) calibrate that model using the relationship between child or child/adult mortality and mortality at other ages; and 3) validate the calibrated model and compare its performance to existing models. To achieve these objectives, the researchers designed a general, parametrizable component model of mortality using singular value decomposition (SVD-Comp). They then calibrated this model using the relationship between child or child/adult mortality and mortality at other ages observed in the Human Mortality Database. Cross-validation was used to validate the model's predictive performance compared to that of the Log-Quad model designed for similar purposes. The results showed that the child-mortality-calibrated SVD-Comp accurately represented observed mortality schedules in the Human Mortality Database. It was also robust against different selections of calibration schedules and outperformed existing models such as Log-Quad Model. The researchers concluded that this approach could be used as a useful tool where child-mortality data is available but information on other age groups' mortalities is unknown. This study builds on earlier work on an HIV prevalence-calibrated version of SVD-Comp, suggesting that this approach is truly general and could be used to develop a wide range of additional useful models.