Two experimental settings in clinical trials: predictive criteria for choosing the sample size in interval estimation