CFE 2019: Start Registration
View Submission - CMStatistics
B1191
Title: Coherent mortality forecasting by weighted multilevel functional principal component approach Authors:  Bo Wang - University of Leicester (United Kingdom) [presenting]
Ruhao Wu - University of Leicester (United Kingdom)
Abstract: In human mortality modelling, if a population consists of several subpopulations it is desirable to model their mortality rates simultaneously while taking into account the heterogeneity among them. Under closely related social, economic and biological backgrounds, mortality patterns of these subpopulations are expected to be non-divergent in the future. In this work we propose a weighted multilevel functional principal component analysis method for coherent mortality modelling, in the sense that the life expectancy in different populations does not diverge in the long run. We treat the mortality rates of subpopulations within a large population as a set of multilevel functional data and use the weighted multilevel functional principal component analysis to extract core information from the functional data and analyse them at multilevel scale, so that the model incorporates both overall information from the population as a whole and specific information from the subpopulations. The proposed model is applied to sex-specific data for nine developed countries, and the results show that the model outperforms some existing models developed in the literature in terms of overall forecasting accuracy.