Figure 5: Platform reproducibility for the Nickel Exposure Study. The diagrams show Lin’s concordance correlation coefficient (CCC), Pearson’s correlation coefficient (PCC), overall Lin’s concordance correlation coefficient (OCCC), and coefficient of variation of replicates averaged across all glycans in array (CV). The CCC and PCC are computed for 28 combination pairs of 8 slides developed with benchmark sera. The OCCC and CV are computed independently for each array: the former is obtained by applying CCC to 66 combination pairs of 12 replicated subarrays from the same PGA. The first diagram (CCC) suggests marginally large bias between slides (20% trimmed mean across all combinations is 0.8), while the PCC shows that the interarray concordance is satisfyingly good if we ignore the inter-array bias, which can be achieved with proper normalization. The last two slides show excellent intra-array concordance, i.e. small measurement error.