STEM Reproducibility by the Numbers

As a scientist, I have concern about the veracity of publications in STEM. I was relieved to read a report1 that provided details on 331 retracted papers in the 1.1 million papers published in the two-year study period (2017–2018). In this interval, the report cites a retraction rate of three in 10,000 papers.

I estimate that a scientist probably reads fewer than 10,000 papers in a long career, so what is the risk and impact? Arguably, the most important figure is the data quality, which includes falsified data plus “problematic” data. The pie chart in Ref. 1 shows that this accounts for less than one-quarter of the 331 retractions. Even if the error rate is a bit more than the metric, 1 in 104 is a very small number. I wonder what the comparable rate is in newspapers such as The New York Times.

More than half of the retractions are attributed to unclear authorship, affiliation and missing author, self-plagiarism, and duplicate publication. These are important defects, but do not directly impinge on the validity and usability of the reported science.

I can see the problem for the non-STEM sciences, which have to deal with the foibles of humans and animals, where cohorts of thousands are large. Chemists are studying large numbers (Avogadro’s number is 6 × 1023 molecules). Plus, none of these exercise free will.

So, when we hear attacks on the lack of reproducibility in science, we can respond with a thoughtful defense based upon our choice of subjects and the tools we use.

With a smile,

Bob

Reference

  1. Chawla, D.S. Plagiarism is top cause of chemistry paper retractions. C&EN June 10, 2019, 97(23), 16.

Robert L. Stevenson, Ph.D., is Editor Emeritus, American Laboratory/Labcompare; e-mail: [email protected]

Related Products

Comments