Confidence distributions for predictive tail probabilities

In this short paper we propose the use of a calibration procedure in order to obtain predictive probabilities for a future random variable of interest. The new calibration method gives rise to a confidence distribution function which probabilities are close to the nominal ones to a high order of approximation. Moreover, the proposed predictive distribution can be easily obtained by means of a bootstrap simulation procedure. A simulation study is presented in order to assess the good properties of our proposal. The calibrated procedure is also applied to a series of real data related to sport records, with the aim of closely estimate the probability of future records.