Inspecting cause-specific mortality curves by simplicial functional data analysis

Cause-specific mortality rates describe the structure of mortality for a particular country and year. In this work we propose to analyze the dynamics of the rates through years and among countries using Functional Data Analysis (FDA). The particular nature of the data – compositions of a whole – requires nontrivial modifications of standard FDA procedures. Specifically, we applied a simplicial FPCA to data in order to show the main modes of variation. Preliminary results about a clustering are also discussed.