17:35
@Oddthinking The criticism I've read says this.
@Oddthinking They do. They try to do it all the time to alleged white racists. They like to see what happens when a supposedly accurate test tells them they're 10% black.
Producers don't care if any of it is real or not. They just want some cheap TV that turns a profit.
Comment from the question:
My understanding from the claimants and critical sources is that they simply compare your DNA to a database. The issue today with accuracy in these tests has nothing to do with the quality of the test, technology, methods, etc. It's the database they are comparing to. The overwhelming majority of the databased entries are white Americans and Europeans. The conclusions I've read say that it's accurate if you happen to be white and the results return "white" ancestry. Otherwise, take it with a grain of salt. —
fredsbend 5 mins ago
So to put it in your words, @Oddthinking, for a person that knows his family has been in Iceland for a thousand years, he can be nearly 100% certain that the "1.5% Polynesian" they claim he's got is nonsense.
Whether the bulk of the numbers come out "80% Nordic, 20% Celtic", I'm not sure, but I bet yes, unless they have "Icelandic" as a unique marker, which I doubt, considering they are only about 1 million people.
Now if you had a Polynesian take the test, My understanding is that he might get "60% Chinese, 20% Thai, 20% Nordic."
Basically, the Chinese and Thai is probably right, meaning, he's got some of that, but the Nordic is nonsense. He has 0% Nordic. But notice that this error is 10 times bigger than the error with the Icelandic guy. This is due to the database issue I mentioned above.
I assume that with time, this problem gets better, as more non-whites are added to the database. Eventually, I suspect they will tell all testers that anything under 10% is x% likely to be a comparative error, and therefore likely not true.
So, on pragmatism, most people will not learn anything they didn't already know. There will be no confidence in "secret remnants" of some distant ancestry.
Most of this I got from an article I read about a year ago. I want to say it was ARS Technica.
Perhaps that was an article on "disease screening":
> In 23andMe’s results table, when I mouse over the section on Alzheimer’s, there’s a big pop-up warning: “NOTE: This result applies to people of European ancestry. We cannot yet estimate risk for those with multiple ancestries.”
This warning should be expected if the database issue I mentioned above is true.
But this article is from 2014. Maybe it's better now.
> Most genetic research studies intentionally limit enrollment to a single population—usually Northern Europeans—since the analysis is easier to carry out in groups tracing ancestry to just one continental region.
That makes sense. If you know of an intercontinental population, it's harder to draw the lines where they mixed and where they didn't. N. Africa and the Middle East would be an example of a population that would have this problem.
It's kind of a catch 22. You can't know about the people you don't know unless you actually do know about enough people first.
But how many people know their heritage 1000 years back? Heck, I don't even know 200 years back.
I have great-grand parents on either side repeating what they heard when they were kids. "We're x% this, x% that, and a little of this other thing."
Since there's no family stories of "Married an Asian" or whatever, I'm pretty confident there isn't any Asian in me. But I bet 23 and me would give me "1.5% Polynesian" or something similar.
