Skip to main content
Fig. 1 | BMC Medical Research Methodology

Fig. 1

From: Random survival forests for dynamic predictions of a time-to-event outcome using a longitudinal biomarker

Fig. 1

Simulation Results. Simulation estimates for AUC (upper panels) and BS (lower panels) for predicted probability P(T∗≤τ+5∣T∗>τ,y(τ),x) from a Cox landmarking model (Cox), joint model (JM), and RSF landmarking approach (RSF). Left panels: a linear trend in marker (Scenario I), middle panels: quadratic trend in marker with non proportional hazards (Scenario II), right panels: Inclusion of two longitudinal covariates (Scenario III, joint model shown is the multivariate model). The lighter “noise” lines represent the models including the additional 7 simulated variables that were not used in generating longitudinal and survival data

Back to article page

BMC Medical Research Methodology

ISSN: 1471-2288

Contact us