Reproducibility probability estimation for testing statistical hypotheses

The applications of statistical tests often generate problems, because of their randomness. Here, stemming from the variability of statistical tests, the reproducibility probability of a statistically significant result (RP) is estimated first. Then, statistical tests are defined on the basis of RP estimation. Lower bounds for RP, and then pointwise RP estimates are used. It is emphasized that the threshold for the RP pointwise estimator to define statistical tests turns out to be 1/2. Hence, taking the perspective of RP estimation, one can compute statistical tests and have suitable interpretations of their results. (c) 2007 Elsevier B.V. All rights reserved.