Thread by @GidMK

Haven't done this in a little while, so let's have a spot of fun

Here's a headline in the Daily Mail. It seems...unlikely that this is true

Let's have a look and see

Lest you think I'm being unfair at the dumb headline, here are the next few paragraphs

Finger length impacts your choice of "masculine" or "feminine" foods! Science?

All credit to the Daily Mail here, they do actually link to the study. Wonderful doi.org/10.1016/j.food…

To the Mail's credit, they appear to have (mostly) fairly represented what the study is about, although they've confused the SIZE of the fingers with the RATIO between length

So, what about the study?

At first glance, I'm pretty unimpressed here. A 216-person online survey with paid participants does not often great science make

This is particularly true when you consider that digit ratios were calculated by the participants themselves. I'd be very concerned about measurement error here...

The sample of "masculine" and "feminine" food combinations consisted of just 5 examples, which is a pretty tiny sample, and the distinction between the two was very modest (1 = totally feminine, 7 = totally masculine)

Then we get to the good (painful?) bit, where the authors dichotomized their continuous variables to make them easier to analyze

This is, uh, not best practice statistically

So, that's not great, but on to the main result!

Digit ratio had no impact on "masculine" vs "feminine" food choice

Um

Oh right, but a subgroup analysis showed a result!

If we only look at hungry participants, those with a "masculine" digit ratio chose more "masculine" food choices (p=0.045)

Here's the graph from the paper. So the red and blue bars on the right are statistically significantly different (p=0.045) whereas none of the other bars are different to each other (they checked)

It's also a touch worrying that the authors appear not to have controlled for multiple comparisons in their analysis, especially given this very marginally significant result

Could've just been luck of the draw!

On that note, it's also a bit odd that there doesn't appear to be a statistical analysis section in the paper. Like, at all. I can't even find what statistical software they used to perform these analyses!

It's also worth noting that there is perhaps another reason that people who were hungry would go for the more "masculine" food option (salad vs burger???)

So the study has...problems. Quite a few of them. These differences are, at best, very modest, and even the distinction between "masculine" and "feminine" foods is...not great

Moreover, I have to wonder whether the sample - 216 Chinese adults recruited/paid through WeChat - are generalizable to many other populations

Going back to the headline, I think we would all look on it a bit more dubiously than we did before

https://twitter.com/GidMK/status/1356429289768382466?s=20