ABSTRACT
Proportional hazards models are widely used in modelling survival data in medical research and other application areas. This assumption of proportionality means that the eect of an explanatory variable or factor on the hazard of an event occurring does not depend on time. This can be quite a strong assumption, and situations frequently arise where this assumption is untenable. Although models that do not assume proportional hazards have been described in earlier chapters, some additional methods are described in this chapter. A particularly important application of non-proportional hazards modelling is the comparison of survival rates between institutions, and so methods that can be used in this context are also described.
