Coherent forecasting of multiple-decrement life tables: Compositional Data Analysis applied to French cause of death data, 1955–2008

This presentation is an introduction to Compositional Data Analysis (CoDA), which should be used when observations have a constant sum, such as percentages or proportions.  A demographic example of Compositional Data is the life-table density of deaths, d(x), which always sums to the life-table radix, independent of the level of mortality. The constant sum also applies to life-tables with d(x) classified by cause-of-death j and age, d(x, j).  The Lee-Carter model is translated into the CoDA form and extended to model and forecast multiple-decrement life tables for France. The principal advantage is that CoDA models the interactions between the competing causes.

• Register here