Estimation of the survival probabilities by adjusting a Cox model to the tail

Within the framework of a survival analysis model with censored life time data and an explanatory covariate, our goal is to predict the survival probability beyond the largest observed time. A Cox model with a constant underlying hazard function is proposed to adjust the tail of the life time distribution. Under some regularity conditions, we prove that the parameter estimators are convergent.