Fig. 3
From: Adaptive sample size determination for the development of clinical prediction models
Comparison of average performance of models based on bootstrap-correction versus the use of a holdout sample for the ovarian cancer data (left) and the CAD data (right)