#somepapers No. 3: Know your data

The paper

Wiedenbeck, M., 2017Proficiency testing: knowing how far you can trust your data. Elements, v. 13, p. 70-72.

What it says

This short paper presents a few case studies to show why proficiency testing is important and some of the common problems that can lead to inaccurate results. Proficiency testing involves sending reference material to multiple labs to evaluate the accuracy (and maybe precision?) of a lab's analytical methods against values believed to represent the actual composition of the material. If there are no problems, you would expect the results from all the labs to form a nice, unskewed Gaussian distribution. It is easier to do this with some elements than others. Problems that can yield inaccurate results include:

• Incomplete dissolution of refractory minerals
• Zircon is hard to dissolve, even in rhyolitic glass, which can lead to underreported Zr and Hf values.
• Assumptions about the isotopic ratios of the material
• The paper discusses a 2.64 Ga pegmatite (high Rb/Sr) that routinely stumps ICP-MS analyses because the instrument commonly measures only ⁸⁸Sr, then corrects for total Sr by normalizing to natural Sr. (I found this case study downright delightful!)
• X-ray self-absorption
• If self-absorption is based on the wrong matrix for the rock being analyzed, the correction will fail. The paper uses an example of Ni concentrations in a rock with high S. Many pressed powder XRF measurements failed because they assumed the Ni would be in silicates (olivine or something, I suppose).

There is a program, GeoPT, that organizes these tests.

Why it matters

An understanding of what can go wrong during analytical work can help design the analytical package used to analyze rocks that help understand whatever system you're investigating.

Why I read it

Honestly, I wasn't planning on reading this paper. I was browsing through the copy of Elements that came in the mail last week, checking out the papers on magma storage in volcanic systems, when I saw this. It was short, so I read through. I don't run a lab (yet?) but I do work a lot with geochemical data from the exploration and production work out at the mine. I am in charge of implementing and reviewing the QA/QC checks for our drilling, so it's important to be reminded of some of the problems that can happen at the lab.

Designing new programs is an important use of this data, but we also have to incorporate historical data with the new data we collect. Understanding what can go wrong at the lab can help us avoid artificial anomalies that can be caused by juxtaposing datasets collected twenty years apart.

Odds and ends

I really enjoyed this short read. I think it could be cause I'm a sucker for case studies. Some of my favorite papers have included a story about how understanding geology (or the analytical side of geochemistry) led to a novel discovery.