|
Abstract:
|
In this thesis we propose an estimator for life expectancy based on the idea of partitioning the condi- tional mean of a random variable into different components. Every component of the Compounded Life Expectancy (CLE) estimator represents a fraction of the population of interest and gives a notion of its contribution to the overall life expectancy. Our approach relies on the correspondence between the cumulative distribution function of a random variable and its quantile function, allowing us to express the conditional mean in terms of conditional quantiles. Every component is related to a certain set of quantiles and therefore to a fraction of our population. A method for quantile regression in the presence of censored data is proposed to estimate the underlying conditional quantile function. Results of two simulation studies show a good performance of the proposed estimator under different scenarios. |
