Fig. 2From: Random survival forests for dynamic predictions of a time-to-event outcome using a longitudinal biomarkerHeart Valve Application Results. Comparison of average cross-validated AUC (left panel) and BS (right panel) of RSF landmarking based on imputation method for missing longitudinal values. Tuned = RSF parameters were tuned; LOCF = imputation using last-observation-carried forward; LMM = imputation using subject-specific prediction from a mixed model for the longitudinal markerBack to article page