Table 8 Duplicates

Approach (see Additional file 4 for example of this approach with real data):

1. Compute precision.

 □ Compute & summarize average relative percent difference (RPD) for duplicate pairs or, if ≥3 side-by-side samples, compute relative standard deviation (RSD):

  ○ If sample results have been normalized (e.g., mass converted to concentration), compute precision with normalized values.

  ○ Compute only for pairs where both samples are detects.

  ○ Also consider precision restricted to pairs where both samples are flagged as “true detects” above the MRL.

2. Visualize duplicate pairs.

 □ This is a good point to pause and check your data and to note/investigate anything that looks unusual (e.g., huge difference in results for two members of a duplicate pair, how tight are detect/non-detect pairs). See Additional file 4: Figure S10A-D for an example.

3. Average duplicates, with non-detects set to lab’s reporting limit.

 □ Calculate average volume and concentration for each pair. Note, can skip this step if only have mass data, or if results were reported by lab as concentrations, rather than as masses that were then normalized to concentrations.

 □ Back calculate new average mass using average volume and average concentration. Or, simply average the duplicate measurements if only have mass data, or if results were reported by lab as concentrations, rather than as masses that were then normalized to concentrations.

 □ Compare new average measurement to MRL to determine data qualifier flag.

 □ Combine duplicate averages back with rest of data.

Reporting:

 □ In publications, we note the range of average RPDs across all chemicals in our QA/QC discussion. We consider average RPD < 30% to be “good” precision.

 □ If a chemical has sporadic blank contamination or variable spike recoveries, excellent precision can increase our confidence in the field sample results.