Fig. 3
From: Weighted Cox regression for the prediction of heterogeneous patient subgroups
Estimated effects in the simulation. Mean estimated regression coefficients of the first 12 prognostic genes in all Cox models for group 1 (averaged across all simulated training data sets, regardless of whether a gene is selected in individual splits or not, and subgroups 1A and 1B) for \(n=50\) and \(\epsilon =0,0.5\). The symbol ‘x’ represents the true simulated effects \(\beta\) of both groups: in black the group of interest (here group 1) and in grey the other group