the sample was "roughly 1,500 young adults 20 to 35 years of age" so, in today's environment, I don't think it's possible that there were not at least a few who self-identify as non-binary and nonconforming.
Transgender folks are represent about half a percent of any given population. So if the study had 1500 participants that would mean ~7 were transgender.
If you look at the graphs from the study those people could be the outliers but no matter what, it's not really enough people to glean anything useful.
> Transgender folks are represent about half a percent of any given population
This is not accurate for the cohort under study. 5% of people under age 30 identify as transgender or nonbinary. I would personally be unsurprised if well over 100 individuals in the study identified that way.
You have to be a little bit careful with those kinds of numbers. Once you get down to a few %, actual signal starts to be swamped by troll answers, misclicks, and other sources of error. You'll basically never see a number below a few percent in any poll with a decent sample size.
(As an aside, this is one of the things that makes the data on the effectiveness of transition care more remarkable than it even looks. Given the very high probability that a random person is cisgender, even moderate false-positive rates would swamp true-positives. The fact that they don't suggests false-positives are very rare indeed.)
What's your point? Looking at their methodology [1], it's pretty clear those sources of error would still be an issue here:
> The American Trends Panel (ATP), created by Pew Research Center, is a nationally representative panel of randomly selected U.S. adults. Panelists participate via self-administered web surveys.
0.5% (as the original poster opined) is not within the 95% confidence interval of the survey. Anywhere as low as 3.2% or as high as 7.8% is within the confidence interval. I am not sure why you would knee-jerk dismiss this.
My point is that it is reasonable to expect between 3% and 8% of the MRI study participants to have been trans. It is not reasonable to think that 0.5% or less would have been trans.
Confidence intervals account for _statistical_ error in which members of the population you happened to survey, not for incorrect data collection like "people randomly select the wrong button sometimes".
I'm not dismissing the claim as false, I'm saying that surveys like this are poor evidence that it is true because they are not very good at making precise statements about rare subgroups.
I don't understand anything you're saying here. It seems to be denying the validity of any survey methodology, no matter how experienced and legitimate the survey-taker is.
Should I doubt that there are really around 5 million religious Jews in the US, because the Pew survey used to estimate this could have had "people randomly selecting the wrong button?" Should I hold it equally likely that there are a mere 500,000 religious Jews? This subgroup is far more rare than the subgroup of 20-35 year olds that identify as transgender or nonbinary. How can we know anything about this subgroup at all, since (according to you) surveys cannot constitute good evidence for anything for a group with such rarity?
The "survey-taker" here was the people being asked. It wasn't administered by a human surveyor; people were completing an online form. I just went and looked at our online signup form, and I've had to collect entry errors on three of the last twenty submissions.
If the only source we had for the percentage of the population that was religiously Jewish was a poll, then yes, I would say you should take that number with a hefty grain of salt. But that isn't the only data we have. We have data from many polls, the census, back-of-the-napkin estimates based on the number of places of worship and average attendance, past demographic knowledge, fertility rates and immigration, and so on.
Even then, there's substantial error in those estimates - the US Census Bureau itself, arguably the most sophisticated demographic operation in the history of mankind, estimates that they double-counted ("erroneous enumeration") roughly ten million Americans, shrugged and went "yeah someone probably lives there" ("whole-person imputation") for another eight million, and missed eighteen million [1]. And that's just for total population (they think the error is larger among subgroups).
And numbers for trans people are highly variable (by as much as 1.5 orders of magnitude) among major sources I'm aware of. For example, the Williams Institute [2] finds ~1.5% trans ID among young people (a third of Pew's number), and binary trans people outnumbering nonbinary ones 2:1 (inverting Pew's ratio). The Canadian census shows <1% [3] among young people.
And corroborating sources are harder to get, both because there's a lot of disagreement about who exactly "counts" (Pew uses "transgender" and "nonbinary" as separate categories, the Williams Institute considers the latter a subset of the former) and because corroborating evidence is often only available for subsets of the population (like people who sought medical care related to their identity).
So yeah, I stand by what I said: a single survey does not tell you very much about a rare subgroup, in this or any other context. It is data, but it should be taken with a substantial grain of salt, especially when it comes to a group like trans people whose boundaries are fuzzy and highly controversial even internally. Certainly I don't think the statistical MOE tells even close to the full story, since it doesn't even for MUCH simpler questions (typical polling error in elections is ~twice the MOE on a much more clearly-defined question, for example).
