A New Model Life Table with an Optimal Number of Age Groups

Ibraheem Ahmed , London School of Hygiene and Tropical Medicine
Julio E. Romero-Prieto, London School of Hygiene and Tropical Medicine (LSHTM)

For countries with inadequate vital registration systems accurate measurements of mortality are difficult. In the absence of reliable mortality data, we rely on model life tables to simulate the real mortality experience of these populations. We propose a new model which simplifies and improves upon earlier approaches. This model requires only a small number of key optimised ages to recover the entire mortality schedule and utilise multiple eigenvectors to explain the error term of the model. Variations in these two features provide models that are suitable for in and out of sample predictions. This model is able to correctly synthesise the general pattern of mortality across a range of scenarios and outperforms the existing Log-Quad and SVD-Comp models by a significant margin. Therefore, this new model’s vastly superior veracity in all settings makes it ideal for application in data deficient settings compared with the existing alternatives.

See extended abstract

 Presented in Session 48. Life Table Methods and Decompositions in Mortality Research