The Time-Varying Effect of Causes of Death on Human Longevity: An Application of Functional Regression with Compositional Covariates

Emanuele Giovanni Depaoli , University of Padova - Department of Statistical Sciences
Marco Stefanucci, University of Trieste
Stefano Mazzuco , University of Padova

In this work, the relationship over time between causes of death and longevity is studied. Specifically, we investigate whether variations in the compositions of causes of death can be predictive of life expectancy at birth, considering four compositions of causes of death related to four age classes, namely 0--4, 5--39, 40--64 and 65+. Employing tools from functional and compositional data analysis, we propose a novel functional concurrent regression model with compositional covariates. A penalized estimation procedure for estimating the regression coefficients and for selecting variables is developed. The proposed model has the advantage not only to determine the impact of various causes of death at different age groups on countries' longevity levels, but also to analyze how these effects change over time. The results confirm the important role of neoplasms and cardiovascular diseases emerged in other studies and reveal other contributions not observed yet.

See extended abstract

 Presented in Session 38. Methodological innovations in the study of mortality