A logistic regression model for survival data

The near non-identifiability of one of the parameters in the Generalized Time-Dependent Logistic (GTDL) Survival Model (MacKenzie, 1996, 1997) is discussed. A new canonical 3-parameter logistic model survival model, in which all of the parameters are identifiable, is obtained. A direct connection with Fisher’s Z distribution is established. The properties of this non-PH model are contrasted briefly with Cox’s PH model. The new model is used to investigate survival from lung cancer in a population study.