Functional Regression of Mortality Compositions: An Application to Italian Provinces

Stefano Mazzuco , University of Padua
Marco Stefanucci, University of Rome Tor Vergata
Gaia Bertarelli, University of Venice Ca' Focari

There is growing literature considering causes death as compositional data for forecasting purposes or in regression models. As for the latter, compositions of mortality by cause has been used as covariates of overall mortality. In this work, we propose to consider the compositions of death by cause as response variable of functional regression model, in order to identify what are the determinants of specific compositional patterns of mortality, particularly focusing on mortality at age 40–64 in Italian provinces between 2003 and 2021. The analysis first involves a functional PCA of compositions, which are eventually regressed with a set of province characteristics. Preliminary reuslts show that causes of death structure of compositions is mostly associated with economic and working conditions of the population.

See extended abstract

 Presented in Session 121. Methodological Innovations in Mortality Studies