Empirical EDA Models to Fit and Project Time Series of Age-Specific Mortality Rates

Abstract

This paper illustrates an application of Exploratory Data Analysis methods to inspect, fit and project a time series of age-specific mortality rates. The analysis centers on Norwegian age-specific mortality rates covering the period 1846- 1988 The main emphasis lies on selecting a simple empirical model facilitating to extrapolate the time dimension in order to obtain mortality projections. As three dimensions may underlie the data-age, period and cohort factors--some attention is devoted to disentangle the relative magnitude of these three sources of variation. To circumvent the well known identification problem caused by the trivial relationship period = age + cohort the estimation method used is based on a weighted iterative procedure along stepwise robust estimation methods. The analysis shows that, other than the effects of the Spanish influenza in 1918 and the Second World War—which produced cohort traces in the form of selection and debilitation effects, but subsided by the sixties—no need is apparent to model cohort effects in the projection. A relatively simple age and period model is adequate to perform the projection.