ABSTRACT
Although the proportional hazards model nds widespread applicability in the analysis of survival data, there are relatively few probability distributions for the survival times that can be used with this model. Moreover, the distributions that are available, principally the Weibull and Gompertz distributions, lead to hazard functions that increase or decrease monotonically. A model that encompasses a wider range of survival time distributions is the accelerated failure time model. In circumstances where the proportional hazards assumption is not tenable, models based on this general family may prove to be fruitful. Again, the Weibull distribution may be adopted for the distribution of survival times in the accelerated failure time model, but some other probability distributions are also available. This chapter therefore begins with a brief survey of alternative distributions for survival data, that may be used in conjunction with an accelerated failure time model. The model itself is then considered in detail in Sections 6.3 to 6.6. One other general family of survival models, known as the proportional odds model, may be useful in some circumstances. This model is described in Section 6.7.
