Figure 2

Displays the power for four different validation scenarios plotted against shift in percent of standard deviation: the first for the complete dataset on a 5% significance level; the second for testing on the validation stage only (N = 500 patients) on 5% significance level; the third for testing on the training set (N = 400 patients) with 10% and on the validation set (N = 500) with 5%; and the fourth scenario for testing on a 5% significance level in the training data (N = 400 patients) and in the validation data (N = 500 patients). There is a sacrifice in power using this validation scheme.