In this diploma first the concepts in connection to survival models are shortly explained, as well as an average expected remaining lifespan at the moment of death and the formula to calculate it. Then the first survival model is described -- the Gompertz-Makeham force of mortality is composed of an age dependent Gompertz component and an age independent Makeham component. Parameters for this model are calculated from data of previous years, typically using the method of least squares. The Lee-Carter model is explained next. To get a forecast of the force of mortality from this model, we only need to forecast the one time parameter $k_{s}$. Parameters for this model are calculated using the singular value decomposition on time period data. Afterwards the parameter $k_{s}$ is forecast using ARIMA methods. Lastly the CH function of survival is explained. It consists of a youth to adulthood component and an old to oldest old component. Its parameters are calculated using the least squares method on mortality tables and forecast using lower-order AR methods. Finally a comparison of all three models reveals that the CH function of survival fits the data better and has more accurate forecasts than the other two models.
|
