Figure 13: The bootstrap test for low and high urinary Nickel samples (each of size 18). The test claims a strong statistical significance of the observed combined AUC value of 0.966. The bootstrap statistic used is combined AUC value obtained by multivariate logistic regression applied to three top selected glycans. Selection of glycans in each bootstrap iteration is performed by Wilcoxon ranking. The test is performed with 1000 resampling by permutation and the resulting test p-value is ASL = 0.005. The diagram shows the empirical distribution under the null hypothesis that the observed AUC value is no larger than any other replicated value. The null distribution has two-sided confidence interval CI = [0.757, 0.947], or one-sided upper bound CU = 0.935. As shown, the observed value is beyond both confidence limits. The empirical data is fitted with a Generalized Extreme Value distribution, which has offered the same p-value as the count of replications above the observed value.