A family of transformations for probabilities is considered for the analysis of grouped survival data. Additive and multiplicative models for the hazard function are compared. A method is suggested for estimating the scale on which an additive representation of the hazard, in terms of explanatory variables, is appropriate. Computations are performed using GLIM. A test for departures from the grouped proportional-hazards model is introduced. An example is analyzed by the methods put forward.