An Extreme Value Approach to Age Limit Analysis

Abstract: A fundamental question in aging research is if the human lifespan has reached its maximum, as it has practical implications and affects the sustainability of our societies and health care systems. This thesis uses methods and techniques from extreme value theory to study human lifespan at extreme age, in particular the age at death of the world’s oldest person (WOP) titleholders. Our study is a contribution to aging research for understanding whether extreme value analysis is useful in modeling life at extreme age and if the models used to fit the data predict the existence or lack of a limit to human life length. Both the stationary and non-stationary models were considered for modeling the data. The stationary generalized extreme value (GEV) distribution and stationary generalized Pareto (GP) distribution which were fitted to the data indicated existence of a limit to human lifespan with an upper bound ranging between 123 and 126 years. A better fit was obtained with the non-stationary GEV model with trend in the location parameter, which took into account the trend in the individual's ages over the years. The non-stationary model does not however indicate an existence of a limit to human life length. Assuming that the linear trend in the location parameter holds in the future, a forecast probability of beating the current world's age record (122.45 years) for individuals born between 1902 and 1965 indicated more than 60% chance of observing a new age record for individuals born after 1940.