We submitted a paper investigating trace elements levels in multiple abiotic and biotic matrices using analytical protocols already established. One of the reviewers required us to report analytical uncertainty calculation for every sample analyzed. We did all of the usual QA/QC procedures (duplicated analyses, CRM analyses and report, etc.), despite that, he still said the analytical uncertainty should be reported for every sample and element analyzed and stated that it is the standard in the majority of studies of our area. Anyone here knows any example of papers doing this calculation and reporting the uncertainty associated with each result?
I am no expert on this specifically, but generally one has an idea how good the data is, and reports in data should include the number of significant figures. If your tables of various elements were always reported to four decimal places with 3-4 significant figures, researchers are going to question What Gives. The table or matrix looks great, but are all those figures significant, or just eyewash off of a computer page. Likewise, if your value reported is 1.025, you uncertainty level is not plus or minus 1, but closer to 0.005, such as you might report when weighing on an analytical balance. Often equipment has reported levels of uncertainty, and when one mixes standards for trace elements and runs them, there comes a point where equipment noise and ability to have uniform linear or curvilinear response with concentration change suggests you are reaching or approaching analytical uncertainty.
Continuing, you might report method and equipment uncertainty in methods, or perhaps a footnote when all samples are generally the same. But for example, your pH uncertainty is going to be greater in a low buffered water as compared with one with high alkalinity, so you might report the former to the nearest 0.1 pH (plus/minus 0.2) and the latter to 0.02 pH (plus or minus 0.05). When you put the pH electrode in the sample, is the value stable or variable. This can help you know what to report. If you run duplicates, is the value the exact same or not. These are the types of things one might use, and some of it may be based on analytical experience.
Dear Gabriel,
haven’t you thought about Monte Carlo simulation?
Uncertainty calculation using Monte Carlo simulation is possible. All uncertainty distributions that are defined in the flows, parameters and characterisation factors are taken into account for the simulation, except the one from the reference product of the system. Best regards, Viktoria
I don't see that request as a standard procedure in most studies of our field. Perhaps the reviewer could provide some reference about what he is expecting? I believe it is worth discussing with the editor.
Good luck!
- Badges
- Science topic