Fig. 2: Prediction scores in the upper 50th percentile of unreliability (top 50% of \(U(\mathop{x}\limits^{\rightharpoonup} )\) values) have worse performance. | npj Digital Medicine

Fig. 2: Prediction scores in the upper 50th percentile of unreliability (top 50% of \(U(\mathop{x}\limits^{\rightharpoonup} )\) values) have worse performance.

From: Identifying unreliable predictions in clinical risk models

Fig. 2

a Calibration curves, b Normalized Brier scores (inset shows expanded region corresponding to \(0.89 \le {\mathrm {Brier}}/{\mathrm {Brier}}_{{\mathrm {null}}} \le 0.912\)), and c Average AUCs for predicting the outcome of all-cause death within 6 months of presentation for predictions in the upper 50th percentile (red) and those within the lower 50th percentile (black). *p < 0.05; **p << 0.0001. Error bars are from 100 bootstrap splits and show one standard deviation (a) or one standard error of the mean (b and c).

Back to article page