B1342
Title: Stochastic modelling and forecasting of mortality rates using a combination of semi-parametric and parametric models
Authors: Erengul Dodd - University of Southampton (United Kingdom) [presenting]
Jon Forster - University of Southampton (United Kingdom)
Peter W F Smith - University of Southampton - Southampton Statistical Sciences Research Institute (United Kingdom)
Jakub Bijak - University of Southampton (United Kingdom)
Abstract: S methodology is described for smoothing and forecasting mortality rates using a combination of generalised additive models (GAMs) and low-dimensional parametric models. GAMs are a flexible class of semi-parametric statistical models, and they allow us to differentially smooth model components (e.g. cohorts) in an integrated way. GAMs can provide a reasonable fit for the ages where there is adequate data. However, estimation and extrapolation of mortality rates using a GAM at higher ages can be problematic due to high variations in crude rates. At these ages, where exposure numbers are small and data are sparse, parametric models can enable a borrowing of strength across age groups and give a more robust fit. Our methodology assumes a smooth transition between a GAM at lower ages and a fully parametric model at higher ages, and acknowledges uncertainty, especially in the highest age groups.
