Fig. 5From: Integrating multiple molecular sources into a clinical risk prediction signature by extracting complementary informationVariability of the.632+ prediction error estimates in the second AML application based on varying overlap sizes of 166, 100, 50 and 30 biological samples, respectively. Integrated prediction error curve estimates for the Cox model, the reference approach and the sequential complementary strategy (verification example). The performance of the Kaplan-Meier benchmark is indicated by a horizontal lineBack to article page