We investigate mortality of the oldest-old by means of nonparametric hazard function estimation based on smoothing techniques for lifetables. Two- dimensional smoothing methods for Lexis diagrams will be developed allowing to predict the survival of demographic cohorts, employing kernel and locally weighted least squares smoothers. Male-female comparisons and changes in the level of mortality will be investigated with change-point techniques. Stochastic process models for samples of lifetables and associated inference will be introduced. It is planned to analyze various biological and demographic data sets with these methods, in particular a huge set of data on the survival of medflies. In this interdisciplinary project, newly developed nonparametric statistical techniques will be applied to approach a complex of biological and demographic qu estions about longevity and mortality of the oldest segment of the population. These questions focus on whether mortality is decreasing for the oldest segment or is invariably increasing with age, how sex differences affect changes in mortality, and to what degree one can predict future mortality and life expectation from current trends. The importance of these problems derives from their potential impact on social planning for the future as well as on the biology of aging. Attacking these problems requires the development of new statistical methods and the innovative application of existing statistical methods to analyze current biological experiments and demographic data
