Did you know that 73.4% of statistics are made up?

I’ve run across several posts all referencing a recent journal article, Retractions in the scientific literature: do authors deliberately commit research fraud?, which parrot (and possibly distort) the information in the abstract — that US researchers are the worst purveyors of fraud, such as this article: US Scientists Significantly More Likely to Publish Fake Research, Study Finds

I wanted to check on that, because it’s not an unknown phenomenon for a article to incorrectly summarize research and so I looked at the article, linked above, to see the abstract

All 788 English language research papers retracted from the PubMed database between 2000 and 2010 were evaluated.

Well, that’s a bit of bias, since people in the US are more likely to publish in English-language journals, but that’s not necessarily true for countries where English is not the native language. It also assumes all fraud is caught and results in a retraction. But beyond that I wanted numbers to look at, since I know there are a lot of articles published in the US, and if they are simply saying that there are more fraudulent articles published in the US, it is pretty meaningless. While I don’t have access to the journal, it turns out that an analysis has already been done. US scientists “more prone” to fake research? No., with some followup in Rates of Scientific Fraud Retractions

The likelihood of a given paper being retracted as fraudulent is higher for China and India than for the US, and significantly so. The finding that the fraud rate is higher is higher-impact journals may be due to having more scrutiny and that we’re simply missing fraud in journals that are not widely read.

I think it’s also important to note (as the paper’s author does) that the overall rate of fraud is low. Using these criteria, it is less than 200 cases out of more than six million papers, or 0.0032%. In other words, for every 31,000 journal articles you read (from all sources), on average one of them will be fraudulent. If you limit yourself to US authors, the number drops to one in only 21,600.