This thesis compares nine different parametric models for survival analysis. Specifically, we compare the Poison, negative Binomial and generalised Poison distributions in combination with three different parametric functions to model the mortality hazard. The comparison is done graphically, with five statistical tests for goodness of fit, and using the information criteria AI, AICC and BIC. For the considered dataset, our results show that the generalised Poison distribution performs the best, followed closely by the negative Binomial distribution. Furthermore, the best parametric function is the Makeham model, followed by the Gompertz and Perks models.
«
This thesis compares nine different parametric models for survival analysis. Specifically, we compare the Poison, negative Binomial and generalised Poison distributions in combination with three different parametric functions to model the mortality hazard. The comparison is done graphically, with five statistical tests for goodness of fit, and using the information criteria AI, AICC and BIC. For the considered dataset, our results show that the generalised Poison distribution performs the best,...
»