M2. Selecting the best model   M2. Effect of the sample size in cross-validation