Support more videos like this at patreon.com/rebecca!
This just in, “researchers say gene changes show who’s gay!” What a great technically accurate but absolutely incorrect headline, NBC!
It’s accurate because some researchers did say that. It’s incorrect because it implies that they said that with any sort of scientific backing, which they did not. It would be like a headline saying “Neil Degrasse Tyson says that Jupiter is made of orange marmalade”. Yes, he did say that, but he was tripping balls at the time and followed it up with “Naaaaaaah.”
The news came not from a published, peer-reviewed study, but from a conference where the researchers talked about their results. That right there should be the end of the story. Until the data has been vetted, there’s just no point in even talking about it. But the conference organizers sent out a press release about it, which was picked up in various places and then eventually landed in the mainstream news. That’s how science news happens these days, unfortunately.
So what does the data look like? Well, it involved 47 pairs of male twins, 37 of whom had one gay brother and one straight brother. The other ten were all gay. So already, we have a really, really small sample size, but not necessarily too small to do a preliminary study. Unfortunately, the researchers then had to split the twins into two groups: a training set to create an algorithm and a testing set to test that algorithm. That means that they were now looking at a sad group of only 23 sets.
23 data points would be bad for most studies, but for this one it’s particularly egregious considering that they were looking at literally thousands of possible characteristics that might correlate. Thousands! And in the end, the model the researchers landed on only managed to predict whether the person was gay 67% of the time.
If those 23 data points were people hanging out at a party, there’s a 50% chance that two of them would share a birthday, and there are only 365 possibilities there. So what do you think the chances are that 67% of them would share one of thousands of possible epigenetic markers?
If you guessed “pretty god damn good,” congratulations! You could be a science writer for NBC News.