Good timing, this issue came up in another thread recently. One thing I noticed about this study is that though the datasets analysed are large, they still suffer from the old problem of self-selection and self-reporting. The primary dataset (Channel 4) is of survey responses from people who watched or heard about a TV programme on autism in 2017. It relies on self-reporting of autistic status. Even assuming the self-reporting is entirely accurate (before diagnosis, I would have self-reported my autism status incorrectly) this is unlikely to be a representative sample of the population, so the results cannot be safely generalised (but they will be by journalists with no knowledge of statistics or how to read a study). The authors acknowledge that this is a problem but brush it aside, repeating that their datasets are large, and citing studies also finding a genderspecial/autistic-traits link but relying on parent- and teacher-reporting of autism status. But parents and teachers are not usually clinical psychologists specialising in autism, a condition we are still just finding the outlines of and about which there are many misconceptions. Just ask any autistic how many people have told them “You don’t look autistic”, or “We’re all a little bit autistic”.

The main problem is that the authors accept the concept of gender identity uncritically. I don’t have a gender identity: there is no mini-me of any gender sitting in my head working the controls. Nevertheless, I would have been shoehorned into the cisgender male category of the main dataset, because I would have stated my sex as male [edit: given my current gender skepticism].