GRADUATION OF MORTALITY DATA USING KERNEL ESTIMATES

Kernel techniques are a non-parametric method that can be used either for density estimation purposes or for regression analysis. Kernel estimates became popular because it is an easy way for representing the structure of a data set. Regarding kernel regression estimates several alternative estimates have been proposed in the literature such as Nadaraya-Watson estimate, Gasser-Muller estimate and kernel weighted local linear estimates. This paper provides a critical presentation and an evaluation of the various kernel regression estimators for graduating age specific mortality data. For that, we apply the alternative kernel estimators to empirical mortality data sets of different national populations and different time periods. 1