In survival analysis, statistical models are frequently specified via the hazard function α( t). A simple model for the relation between the hazard functions in two groups (e.g., a treatment group 1 and a control group 0) is the proportional hazards model where \(\displaystyle \alpha _1 (t) = \theta \alpha _0 (t), \) and θ is the treatment effect. For a parametrically specified baseline hazard, α 0( t), both the treatment effect and the parameters in the baseline hazard are usually estimated using maximum likelihood. In a semi-parametric model where the baseline hazard is left unspecified several estimators for θ are available: the maximum partial likelihood estimator, cf. Cox (J R Stat Soc Ser B 34:187–220, 1972), a class of rank estimators, and some ad hoc estimators.