Smoothing Methods for the Analysis of Mortality Development

This is the title of my PhD dissertation which I developed under the supervision of Jutta Gampe from the Max Planck Institute for Demographic Research and Maria L. Durban Reguera from the Universidad Carlos III de Madrid. Devoted to the analysis of mortality development on the aggregate level, in this dissertation I used smoothing methods, in particular to achieve appropriate measures of fit for large samples and studied models based on warping age-at-death distributions as well as modeling digit preferences via smooth latent distributions.

Concerning the first objective, I suggested a novel and hopefully more appropriate statistical measures of variability that allow comparison of different smooth mortality models in the presence of (very) large sample sizes. My second objective was to consider models that operate on the probability density function of the lifetime, i.e. the age-at-death distribution, instead of the more common consideration of the hazard function. The third objective concerned a peculiar source of overdispersion in mortality data, the so-called digit preference.

My PhD dissertation is freely available at the following link: PhD thesis