Loyal readers of this blog may have been wondering if there have been any updates on the Nicolas Guéguen story since I wrote this post back in June 2020. Well, actually there have!
First, in April 2022 an Expression of Concern was issued regarding the article that I discussed in this open letter, which was the first paper by Guéguen that I ever looked at. Of course, issuing an EOC — which is still in place over two years later and will probably last until the heat death of the universe — is completely absurd, given that we have smoking-gun level evidence of fraud in this case, but I suppose we have to be grateful for small mercies in this business. Guéguen now has 3 retractions and 10 expressions of concern. Hallelujah, I guess.
Second, after a hiatus of about 7 years since James Heathers and I first started investigating his work, Guéguen has started publishing again! With co-authors, to be sure (almost all of our critiques so far have been of his solo-authored papers, which makes things less messy), but nevertheless, he's back in business. Will it be a solid registered report with open data, fit for the brave new post-train wreck world, or will it be more Benny Hill Science™? Let's take a look:
On 2024-08-08 I was able to download the article from here.
I think it's fair to say that the European Journal of Public Health Studies is not what most people would regard as a top-drawer journal. It does not appear to be indexed in Scopus or PubMed and its website is rather modest. On the other hand, its article processing charge is just $85, which is hard to argue with, and of course it's what's in the paper that counts.
The study involved finding ways to get children to eat more fruits and/or vegetables, which may ring a few bells. I'll let you read the paper to see what each variable means, but basically, children aged 8 or 9 were asked a number of questions on a 1–7 scale about how much they liked, or were likely to consume, fruits or vegetables after a brief intervention (i.e., having an adult talk to the child about their own childhood food preferences — the "Similarity" condition — or not).
Let's have a look at the results. Here is Table 2:
First, note that although the sample size was originally reported as 51 (25 Similarity, 26 Control), and the t tests in Table 1 reflect that with their 49 degrees of freedom, here we have df=48. Visual inspection of the means (you can do GRIM in your head with sufficiently regular numbers), backed up with some calculation because I am getting old, suggests that the only possibility that is consistent with the reported means is that one participant is missing from the control condition, so we can continue on the basis that N=25 in each condition.
There is quite a ceiling effect going on in the Similarity condition. Perhaps this is not unreasonable; these are numbers reported on a 1–7 scale by children, who are presumably mostly eager to help the researcher and might well answer 7 on a social-desirability basis (a factor that the authors do not seem to have taken into account, at least as far as one can tell from reading their "limitations" paragraph). I set out to use SPRITE to see what the pattern of likely individual responses might be, and that was where the fun started. For both "Pleasure to respond" and "Feeling understood by the instructor", SPRITE finds no solution. I also attempted to generate a solution manually for each of those variables, but without success. (If you want to try it yourself, use this spreadsheet. I would love to know if you can get both pink cells to turn green.)
Thus, we have not one but two examples of that rarest of things, a GRIMMER inconsistency, in the same paper. I haven't been this excited since 2016. (OK, GRIMMEST is probably even rarer, although we do have at least one case, also from Guéguen, and I seem to vaguely remember that Brian Wansink may have had one too).
I am about to go on vacation, but when I return I hope to blog about another recent paper from the same author, this time featuring (drum roll please) RIVETS, which I like to think of as the Inverted Jenny of error detection.