Figure 2
From: Allowing for mandatory covariates in boosting estimation of sparse high-dimensional survival models
.632+ prediction error estimates. .632+ prediction error curve estimates for microarray-only models (left panel) fitted by CoxPath, gradient boosting, and componentwise CoxBoost, for models including the IPI as an additional optional covariate (middle panel), and for the CoxBoost fit that incorporates the IPI as a mandatory covariate (right panel). The Kaplan-Meier benchmark is indicated by gray curves.