Skip to main content
. 2015 Oct 29;35(7):1159–1177. doi: 10.1002/sim.6782

Table 3.

External validation of the models for the probability of a thromboembolic event.

Performance measure Calibration slope (s.e.) C‐statistic (s.e.) Brier score
MLE 0.78 (0.12) 0.704 (0.025) 0.03396
BE 0.85 (0.13) 0.703 (0.025) 0.03381
RIDGE 1.27 (0.18) 0.724 (0.024) 0.03357
LASSO 1.23 (0.17) 0.718 (0.024) 0.03359
ENET 1.22 (0.17) 0.723 (0.024) 0.03358
ALASSO 1.07 (0.16) 0.715 (0.026) 0.03361
SCAD 1.07 (0.16) 0.697 (0.026) 0.03370
BLASSO 1.26 (0.17) 0.723 (0.024) 0.03360
SSVS 1.19 (0.17) 0.718 (0.024) 0.03359

Performance measures were evaluated on a validation data set with 2739 patients and 97 events. MLE, maximum likelihood estimation; BE, backwards elimination; LSF, linear shrinkage factor; ENET, elastic net; ALASSO, adaptive lasso; BLASSO, Bayesian lasso; SSVS, stochastic search variable selection; SCAD, smoothly clipped absolute deviation.