Inference on the probability P(T1<T2)P(T1<T2) as a measurement of treatment effect under a density ratio model and random censoring

Let T1T1 and T2T2 be the survival times of patients randomized to, respectively, two treatment groups. The probability P(T1