How to Get a Robust Estimation of the Modal Length of Life?

Siu Lan Karen Cheung, Institut National de la Santé et de la Recherche Médicale (INSERM)
Graziella Caselli, Università di Roma "La Sapienza"

Demographers and bio-medical scientists have been concerned with life durations. Instead of focusing on life expectancy at birth with primarily dependent on infant and premature deaths or any arbitrary selection of age limits in term of age-specific death rates, some researchers have proposed using the modal length of life in the distribution of age at death as the best indicator of the human life span which gives a good account of old age mortality. Data from low mortality countries show a certain flatness and large confidence interval in the curve near the region of the mode causing the mode being sensitive to minor variations which makes it important that the curve is unimodal and relatively smooth. Previous simple smoothing techniques lack of potential statistical feedbacks. For this case, our results show that a quadratic fitting in the region allows a robust estimation of the mode and discloses trends over time.