Title: Joint graduation of male and female mortality rates using penalised splines
Authors: Kai Hon Tang - University of Southampton (United Kingdom) [presenting]
Erengul Dodd - (United Kingdom)
Jon Forster - University of Southampton (United Kingdom)
Abstract: Splines with an exponentially increasing penalty are applied to period mortality data, creating a flexible yet robust method for graduation of mortality rates. Previous parametric graduation methods usually involve complicated mathematical functions due to the peculiar mortality pattern, the estimated parameters are often highly correlated hence introducing difficulties during the fitting process. By utilising P-splines, the model enjoys computational efficiency and flexibility. Observed data at high ages are often sparse and unreliable, an additional difference penalty between male and female spline coefficients is included to further strengthen the robustness and borrow information from each other. Constrained splines are used to prevent the cross-over of male and female mortality patterns when they are modelled jointly.