Why determining ancestry is rarely accurate

[ganzfeld]

Of course they generally use tech better suited for identifying whether a given person is likely to be a parent, which is a different problem. I would submit some do use better tech and some don't. That's why (unlike DNA ancestry) there's a accreditation process. Not all parentage vendors are accredited. (Also the price is not orders of magnitude different. Depends on the company and the test.)

In any case it's more like comparing one telescope to see the planets and a different one to see the moon. Sure, you get better pics of the moon but it's not necessarily because the scope is better.

[erwins]

Originally Posted by jimmy101_again

Yes. Indeed if you read the legal literature nad court cases the ability to prove non-parent hood is generally considered to be scientifically irrefutable. It can only be refuted on procedural grounds but EVERYTHING in a legal case is refutable on those grounds.

The reliability to prove parenthood is almost as strong and I would argue that even that is stronger than virtually anything else in a trial. Eyewitnesses are notoriously unreliable. Confessions are notorious unreliable. Even finger prints are no longer considered to be foolproof, like they once were, because there is a fair amount of subjectivity in fingerprint matching.
Fingerprint failure rate roughly 0.1% in this study http://www.pnas.org/content/108/19/7733

Also, https://www.pbs.org/wgbh/frontline/a...so-scientific/
Teeth marks, hair, even fingerprints are considered less reliable than DNA.

I tried clicking on the underlined "scientifically irrefutable," but you seem to have forgotten to link your cite there.

[DawnStorm]

Originally Posted by NobleHunter

I'm pleased to know these companies are full of it.

Being that part of my family is Irish Catholic, even primary records aren't too useful. It's really tough telling all those Marys apart.

Well there's Proud Mary, the Virgin Mary, Mother Mary, Cross-eyed Mary; Mary with her little lamb; the Quite Contrary one, et al.

My paternal uncle was way into genealogy long before it was cool. He gave me parents this thick green book that just listed names and dates. The most interesting part was the page where my name was listed. Other than that, it was a real eye-glazer. I do know that my dad's ancestors hailed from England, Wales and Scotland and my mom's ancestors were German. So I'm a Scots-Irish woman waving a German flag. Must explain why I like Dobermans.

[Beachlife!]

I've recently had a DNA test done, and just signed up for a second. In the process I also signed up for Ancestry.com, for one month.

Ancestry allows each person to build their own family tree. It provides access to a whole lot of supporting documents. These are all legal documents which will give one a half dozen or more generations of family history. That's enough to trace several of my ancestors to 'the boat'. From my own stand point that's enough.

The DNA test I had done tells me about my 'deep ancestry' which compares my DNA that to people in general regions of the world. In my case all my ancestors fall into four regions of Europe, the smallest of which accounts for 5%. I don't know what the smallest percentage they show would be or if there is a maximum number of regions they will show for an individual.

The test also shows a trace of some of my DNA markers which follow a path all the way back to human origins in Africa. Although I believe the information is accurate, its not clear to me why there are only two of these markers. Essentially, I am seeing a trace of two 'random' ancestors.

[Alarm]

Originally Posted by DawnStorm

Well there's Proud Mary, the Virgin Mary, Mother Mary, Cross-eyed Mary; Mary with her little lamb; the Quite Contrary one, et al.

My paternal uncle was way into genealogy long before it was cool. He gave me parents this thick green book that just listed names and dates. The most interesting part was the page where my name was listed. Other than that, it was a real eye-glazer. I do know that my dad's ancestors hailed from England, Wales and Scotland and my mom's ancestors were German. So I'm a Scots-Irish woman waving a German flag. Must explain why I like Dobermans.

My ancestors are from France... Poitiers (circa 1614) on one side and Berry/Guyenne (circa 1668) on the other

[Errata]

I got one of these for Christmas, since apparently I'm hard to shop for, and just got it back. Yes it means they're going to sell my clones to the military industrial complex, and show me genetically specific pop-up ads, but whatever.

Originally Posted by Esprise Me

I've been surprised by how many of my clients have already shelled out for one of these tests, or are eager to do so. I understand they're pretty expensive, and in any case, getting a result that you have Native American heritage doesn't mean your child is eligible for membership in a tribe. But in any event, I've yet to meet someone who took one of those tests and didn't get a positive result for Native American.

The further back you go, your ancestors grow exponentially, so it's plausible that lots of Americans really do have trace amount of Native American ancestry. But count me as someone who didn't. I got 0% Native American. 98% European and 2% Middle Eastern. Which is mildly surprising, since the branch of my family that I have the most complete information on came to the US in the 1600s from Wales, and some lived in early Caribbean colonies. That's 10+ generations, which could mean hundreds of ancestors in the Americas. It seems very likely that some of them could have been black or Native American in all that time, but apparently not.

The bulk of my ancestry was entirely unsurprising. 75% Irish, Welsh, or Scottish. However the smaller components were not what I would have guessed, or easily explained. I got around 1/8 Scandinavian, and 1/8 various Mediterranean, which is almost exactly a full great grandparent each, but I can't reconcile either with any of my great grandparents. And the other surprise is 0% French. One of my great great grandparents was supposed to be French by way of Louisiana. Maybe she was just culturally French. Or someone passed on incorrect information about her history or someone's parentage. Or it's just the margin of error and the test is not good at figuring out ethnicity.

Regardless of how accurate it may be with these sorts of things, it did correctly identify that I was 4% related to a second cousin, and 26% related to my uncle. So it can at least do that much.

[ganzfeld]

There's no such "ground truth" for any of those gene sets. They would have to get a large sample of high confidence data. In the case of indigenous groups, there are several problems, not least of which there has been some mixing on that side as well. In the case of groups like "French" as opposed to "various Mediterranean" for example, what would be the slice of time in history or prehistory that they even mean? How likely would it be, for example, to be from northern Great Britain (or anywhere else in Northern Europe for that matter) without having any Scandinavian at all? Given the history of those groups, it would be at least rather uncommon. So the stats are a bit nonsensical. All it means is that certain markers they have identified in current populations also exist in the donors, which really means nothing of great interest because they don't even necessarily correspond well to the prevalence of those genes at any other time or place in history. It's a bit better than a palm reading but, IMO, not much.

This is a bit like when they say "what's the most __ " in every state/country and they have a map and they're all unique. Well, it turns out that they had to do a bit of a trick on the way they figured out "most" - when you look at the fine print. But most of these companies don't actually give you the fine print. It's just, trust us, we're DNA experts.