We study families of two-sample tests with grouped or multinomial data. The tests are based on models for which some function of hazard or continuation ratio probabilities in two distributions differ ...