Robust forecasting of mortality and fertility rates: A functional data approach

Spatial pattern and its evolution of Chinese provincial population: Methods and empirical study

“…This method is seldom applied for forecasting population on the national level. (Hyndman and Ullah, 2007). The C-C model divides the population by age groups, and takes effects of human fertility, mortality and natural growth into consideration, thus it provides a relatively accurate prediction of population (Stauffer, 2002).…”

Point and interval forecasts of age-specific life expectancies

“…In these industries, the viability of financial arrangements depends on knowing the likelihood that clients will live to older ages. In the demographic literature, a number of parametric and nonparametric methods have been put forward for forecasting age-specific mortality rates and life expectancy at birth (see for example, Preston, Heuveline, and Guillot 2001;Rowland 2003;Alho and Spencer 2005;Hyndman and Ullah 2007;Torri and Vaupel 2012). In a recent paper by Shang, Booth, and Hyndman (2011), they compared the point and interval forecast accuracy for forecasting age-specific mortality rates and life expectancy at birth, among ten principal component approaches.…”

“…As emphasized by Bell (1997), it is desirable to use modeling and forecasting methods that capture a smooth shape over age for producing consistent forecasts and improving the accuracy of short-term forecasts. To solve this issue, Hyndman and Ullah (2007) combined the ideas of nonparametric penalized regression spline and functional principal component analysis using second or higher order functional principal components.…”

A dynamic parameterization modeling for the age–period–cohort mortality

“…Renshaw and Haberman (2003a) incorporate age differential effects, introducing a double bilinear predictor structure into the LC forecasting methodology, and optimize the Poisson likelihood, as opposed to optimizing the Gaussian likelihood, as under the LC approach, and then compare the results. Also, Hyndman and Ullah (2005) use several PCs in order to capture the differential movements in age-specific mortality rates. They smooth first the observed log-mortality rates with constrained and weighted penalized regression splines and they decompose the fitted curves using functional PCA.…”

“…Thus, Currie et al (2004) use bivariate penalized B-splines to smooth the mortality surface in both the time and age dimensions within a penalized GLM framework. Hyndman and Ullah (2005) smooth the observed log-mortality rates with constrained and weighted penalized regression splines. De Jong and Tickle (2006) introduce a state space framework using B-spline smoothing.…”