Reflections on Alternative Approaches in Causes of Death Data: Cause-Specific Hazard Rates or Compositions by Cause?

Stefano Mazzuco, University of Padova
Marco Stefanucci, University of Rome Tor Vergata

Given the higher availability of relevant data, there is a growing body of literature on causes of death analysis. In later years, there has been increasing interest in using the Compositional Data Analysis (CDA) approach in analysing causes of death data as an alternative to cause-specific hazard rates, which underlie a competing risk model. By means of simulations, the paper shows how the two approaches behave when the dependence across competing risks are mispecified, and it turns out that compositions are less affected than cause specific rates. It should also be noted that analysing causes of death compositions is not equivalent to decomposing mortality measures by cause. In this article, potentialities and issues related to both approaches are considered.

Keywords: Decomposition analysis, Mathematical demography , Mortality and Longevity

See paper.