Skip to main content
Fig. 7 | BMC Medical Research Methodology

Fig. 7

From: Assessment of predictive performance in incomplete data by combining internal validation and multiple imputation

Fig. 7

Type 1 error and power of resampling-based confidence intervals for AUC and ΔAUC estimates. Percentage of rejected null hypotheses (i.e., confidence interval above 0.5 and 0 for AUC (a, b, c, d) and ΔAUC, (e, f, g, h) respectively) among 250 simulations plotted against the underlying true (theoretical) value. In the absence of a true effect (true auc=0.5; Δauc=0), percentage of rejected null hypotheses equals type 1 error, otherwise power. Parameters were chosen as denoted in the figure titles, n=200,p0=0,1 and otherwise as in Fig. 3

Back to article page