Can the Help for Triangulation be updated?

Question

Could we get the help for triangulation updated? There has been some confusion about the wording of the requirements. I have seen many attempts at triangulation using matches closer than 3c1r.

The current help states:

Three or more cousins need to all match each other on a single segment of DNA that is at least 7cM long.

I believe that the confusion stems from the use of the word "cousins" wthout qualification. 

The help following talks about the three-legged stool comparison and tends to further confuse the situation, in my opinion.

Suggestion: Change "Three or more cousins" to "Three or more distant cousins" and delete the three-legged stool comparison. 

---

(Edited to add further clarification to suggestion.)

Suggestion: Also change the previous sentence to clarify what is meant by "distant cousin". Here is how I would rewrite both sentences:

Confirming relationships using matches between cousins more distant than 3rd cousins requires triangulation. Three or more distant cousins need to all match each other on a single segment of DNA that is at least 7cM long.

(Edited to fix error) 

Comments

I added the Leaders tag so that all leaders are aware of your question.

---

Hi Frank,

You're correct that the DNA confirmation guidance here doesn't specify how to confirm when certain specific relationships are involved, one such relationship being the H2C1R that you mentioned.

If you believe that the intent of the DNA confirmation guidance is such that matches with relationships that aren't specifically called out in the guidance shouldn't be used for DNA confirmation here, and that doing so would be "breaking the rules", then you probably shouldn't use such matching relationships for confirming child-parent relationships here.

But I don't believe that is the intent of the DNA confirmation guidance here, so if the DNA confirmation guidance isn't clear, my plan is to take reasonable steps (including using the relationship groupings that I mentioned above) to determine which DNA confirmation process to follow here for a particular match's relationship.

---

Thank you!

---

How about this:

Requirements for Triangulation:

Simple one-to-one DNA confirmation is sufficient if your match is a third cousin or closer. Confirming relationships using matches between cousins more distant than 3rd cousins requires triangulation. Three or more distant cousins need to all match each other on a single segment of DNA that is at least 7cM long. One match may converge on an ancestor more recent than the most recent common ancestor (MRCA) for all three, but this ancestor should not be more than two generations below the MRCA. Seven cM is a bare minimum and presumes that the cousins' relationships back to their MRCA(s) are well-documented.

---

Rick,

I agree. These rules can be difficult to write clearly, and to understand. I assume that they simply didn't think things through to that level of detail - no doubt it was challenging enough to get to where they got, without throwing in additional complications.

I wouldn't have a problem with someone treating a H2C1R exactly the same as a 3C in this context, because I know it's about the amount of shared ancestry. But others might not realize that's the case.

Really, it more about, for example, a H2C2R, which I would argue is "more distant than 3C", while people might just look upon it as a flavor of 2C.

It's a recipe for confusion, and so it'd be better if that aspect of it were spelled out. Better still would be if it were automated, so the user doesn't really have to think about it.

---

John,

That spells things out a bit more. I think some people really like the "stool" visualization, but others just get confused at the introduction of a new and abstract concept. If it can be explained reasonably simply without "the stool", that might be better.

I'd suggest a slight re-organization though, writing-wise: Putting the last sentence, which is talking about the 7cM, immediately after where the 7cM requirement is brought up. Then maybe make the next part a new paragraph.

It also occurred to me in other discussions on this thread that there isn't any reason to exclude 3C or closer, and in fact most 3C matches on AncestryDNA don't meet "Rule 5" for one-to-one confirmation. In those cases, triangulation would be the way to go.

Fundamentally, there's nothing at all about a closer-relation match that means you can't do triangulation, it's just that if you can do a one-to-one instead, that that's inherently the better thing to do.

So I'd suggest an approach that's a bit different:

* First, define what constitutes a triangulation - three DNA tests among people with a common ancestor (or a male-female pair of common ancestors), where they all match each other on the same DNA segment. As it is, they just kind of jump into things without even saying exactly what's being talked about. And they start by literally talking about something ELSE.

* Second, (and I mean in a second paragraph) identify exactly what situations might offer the opportunity to do a one-to-one confirmation instead. Recommend that the reader look into that option, and to use it instead if the requirements for doing so can be met. This is really the only place we need to talk about 3Cs (or equivalents), because triangulation CAN be used for ANY relation - it's just that it's easier to do the one-on-one way when you can.

* Third (and 3rd paragraph), the stuff about needing 7 cM.

* Fourth (last paragraph), explain how the MRCA for each possible pair of matches among the three can only be one (or whatever number is decided) generation below the MRCA for all three.

The single paragraph you have isn't super long, but shorter pieces to "digest", one at a time, should make it a little easier for the reader, especially since I'm talking about adding a little introductory material at the beginning.

---

After reading the triangulation help and the discussion here about the three leg stool, I am a bit confused.

It seems that the requirements for the 3 persons might be defined as:

1. Relationship to each other: They must not be siblings, parent/child, ... (not clear to me what these requirements are)

2. Degrees of separation from the MRCA (or number of ancestors): must not differ by more than one.

---

Paul,

1. All of the testers in the triangulation group must be no closer than 3rd cousins once removed to each other.
2.  The MRCA of each pair of testers should be the same or not separated by more than one generation from the oldest MRCA of the group.

---

Thanks John. That makes sense.

It would be helpful if the owner of the help text could add this clarification.

---

I also think that WT should emphasize that you can apparently have more than three people in a "triangulation" group.  That term, and the use of the stool analogy suggests that you are limited to three people only.

Answer 1

"Cousin" is often used to mean "1st cousin", so maybe that's a good idea. Then again, "distant cousin" means different things to different people. Maybe "relation" would work.

But there's an additional wrinkle to the triangulation requirements that you may be missing, and which is probably in the "three-legged stool" discussion. As I understand it, the MRCA for any two can only be one generation below the MRCA for all three. You could say that one "leg" cannot be much shorter than the other two.

I could not use, for example, myself, my brother, and a 5th cousin. I could not even use myself, a third cousin, and a 5th cousin. But I COULD use myself, a 4th cousin, and a 5th cousin. It'd be best if all three are 5th cousins to each other - each descended from a different child of the MCRAs - but there's a little leeway there.

Comments

I believe that is so in order to be able to triangulate from an ancestor who only had two children.

---

I agree with Frank that removing the three-legged stool section from the Triangulation Help page could be problematic. I believe that my understanding is similar to Frank's - I believe that we're trying to ensure that within a triangulation group, the MRCA for any two matches can only be as far away as one generation below the MRCA for all three (or more) matches. I believe the three-legged stool section is the only place in the Triangulation Help page in which that constraint is addressed.

---

Why is the one generation maximum important? I can understand the stool's example of two first cousins and a fourth cousin not working, but how about 3c1r, 4c1r and 4c2r, for example? One match is two generations down instead of one, but is still 3c1r.

---

I agree that some of these rules are arbitrary in that they eliminate some valid triangulations.  The difficulty is in how to restrict the lineages so that one is certain that no other path is possible. Only having full sets of documented trees allows one to make these judgements.  For endogenous populations, this might be impossible even with fully documented trees and all three distant cousins with the same MRCA.  I would also question why all three must be beyond 3rd cousin... for example: two 3rd cousins validate their lines without triangulation but if they both share a segment with a mutual 3rd cousin once removed, it is not valid in validating that 3rd cousin once removed line according to the the rules. Oddly enough, if one were to replace one of the the two 3rd cousins with a child of theirs sharing that same segment, THAT is valid by the rules. Where is the sense in that?

Long story short, Occam’s Razor does not allow one to remove necessary considerations.

---

Hi John,

This G2G post discussed the proposed changes to the DNA confirmation guidance at the time. Kerry Larson suggested in his answer regarding triangulation "The legs need to come together at about the same level.  Perhaps the branches should be within one or two generations of the MRCAs?" It appears that the guidance verbiage settled upon was "All three legs should meet at the seat, or if not, then no more than one generation below it."

---

Thom,

The way I read the triangulation rules, it DOESN'T preclude the case you're talking about, where the three legs are you, a 3rd cousin, and the child of someone who is a 3rd cousin to the first two.

I think it's just that it's geared towards YOU confirming your own ANCESTORS. But what if you've got quite the family tree input, and you want to use whatever DNA is out there to validate some of the OTHER branches? You certainly can do it, but the documentation isn't necessarily written from that perspective.

In this example case, you'd do the confirmation with the 3rd cousin (if the match is sufficient), and that would confirm back to a pair of gt-gts for you. The 3C1R doesn't add anything.

But a triangulation among the three gives a confirmation for the line from your gt-grandparent's sibling, down to the 3C1R (assuming they're in the generation after you). For you, it only confirms back to your gt-grandparent, which is not as far back as what the confirmation with your 3C gives you.

Actually, if it's an AncestryDNA match, if your match to your 3C is less than 90cM, you get screwed over by Rule 5, and the above triangulation is ALL YOU CAN DO., resulting in confirmation only back as far as your gt-grandparent.

---

The thing about the "one generation difference limit for a leg" is that the whole idea is to mitigate to possibility that the segment isn't really from some 6th-great grandparent you don't even know you have, instead of from the ancestor you know about.

Suppose there's a 1-in-100 chance of my sharing a mystery 6th-gt grandparent with my 5th cousin, and our match coming from that. If I add in my BROTHER as the third leg, he has the same ancestors as I do, so if he has that segment too it basically means ZERO to the confirmation.

But if I use ANOTHER 5th cousin (also a 5th cousin to the first), the chance of both 5th cousins having that mystery ancestor might be 1-in-10000 (100 times 100). So the "third leg" makes it FAR less likely that the mystery extra common ancestor is out there.

The closer any two legs are related, the higher the chance of some mystery ancestor being the REAL source of the common DNA segment. So it's a judgement call, on how many generations down from the MRCA to allow. Maybe they allow the one, mostly because some ancestors only have one child with living descendants.

You could say the same thing about how many legs! It would be FAR less likely for the DNA segment to be from the wrong ancestor if you had FOUR 5th cousins, all descended from a different sibling. But if that was the requirement, it would be even more rare to be able to do it than it already is!

Answer 2

Hi John

Yes, but..

Does it actually do any harm if someone wants to demonstrate triangulation between 3rd cousins? I know it isn't necessary, but does it hurt?

The three legged stool analogy was introduced because the previous wording was far too proscriptive and limited whatever cousins were involved as meeting together at exactly the same MRCA. That is both an unsupported requirement and often impossible to achieve.

At the time I was in favour of extending the "seat" of the stool by two generations rather than one and there have been other G2G comments recently in favour of this. I do not have any rigorous proof either way of this (perhaps Ed does if he is following these questions) - my gut feel is that if the match is sufficiently large then the generation gap could be extended, but I have no feel for how far a given size of match should extend the allowed gap. I would also struggle to put this into words that were easy for non-DNA experts to follow.

Comments

Hi Derrick,

I guess it all depends on how verifiable WikiTree (Chris) wants DNA confirmation sources to be. It certainly does no harm for people to add auDNA triangulations as "suggested" instead of "confirmed" relationships or research notes if the matches don't meet the requirements for DNA confirmation.

Without more automation, I think it is unrealistic to believe that a profile having the the DNA confirmation relationship indicator set via an auDNA triangulation necessarily means that the requirements for DNA confirmation have been met. But that's a topic that might be better discussed under the "accuracy" label.

---

These are good points, and bring to mind what a poor analytical footing we are on. Derrick is getting into "fine tuning", and to reasonably do that you need a better understanding of exactly how things are, and exactly what level of confidence you want to require.

There are some mathematical models, but I'm not sure how great they do. For empirical data they generally look to the "Shared CM Project" stuff, which people seem to have a lot of faith in, because they don't think for themselves - it's a mess. As far as I know, the whole field seems kind of "dead", relatively speaking. Much fewer people getting tests. Few people ever tried to make the most of the tests they got. There's a lot of information to work with, but not enough people with the ability or time to do it, and it takes a lot of co-operation where often little or none is to be had.

So it seems kind of bleak, as far as doing much to advance the state of this reasonably, any time soon. As it is, the whole thing should be an "app" - software that asks you some questions, and handles the bureaucracy and rules for you. I even volunteered for the DNA Project some time ago (well over a year, I'm sure) but you don't see me with a badge for that, do you? Last I checked, the last such badge awarded was even longer ago. It seems like a dead project. DNA became a big thing just a few years ago, and the full benefits still untapped, and it seems dead. It's sad.

I may be out-of-date on this, but there was a LENGTHY thread with many suggestions back then - I don't think much of anything has changed at all, from what I've seen.

---

It occurred to me that sometimes triangulation IS necessary for 3rd cousins. Rule 5 can disqualify a match from being used - AncestryDNA matches under 90cM being quite common (5 out of 6 AncestryDNA 3C matches are below that, in my experience).

But you only get confirmed back to the gt-grandparent, not both gt-gts.

---

"Does it actually do any harm if someone wants to demonstrate triangulation between 3rd cousins? I know it isn't necessary, but does it hurt?"

I have a real case where three 3rd cousins have each taken Y-111 DNA test. The three 3rd cousins have been determined by FTDNA to be a close Y-chromosome match. This is a material fact. It is my opinion, that demonstrating triangulation in this situation discloses this material fact, adds additional confidence, and does no harm.

However, current Wikitree guidelines do not allow triangulation for three 3rd cousins to be included in the compliance statement.

Answer 3

Please consider making the distinction between AUTOSOMAL  DNA and Y-DNA.

I agree - Triangulation on WikiTree is not needed for third cousin matches.

I agree - Triangulation on WikiTree  should not include AUTOSOMAL matches which are third cousins or closer."

I believe - Triangulation on WikiTree  COULD BE and SHOULD BE USED for Y-111 matches which are third cousins or closer.

I used Y-111 DNA triangulation for three third cousins. However, I was requested to, and I did remove the triangulation statements from several profiles, because the Wikitree instructions do not allow Y-111 Triangulation for three third cousins.

This is a real situation. There are three third cousins. Each has taken the Y-111 DNA test. Each of the three third cousins are a Family Tree Y-DNA close match with a genetic distance of (0-zero) and (1-one).

Prior to the Y-111 DNA test, none of the three third cousins knew that the other two third cousins existed.  Two of the third cousins only knew who their respective grandmother and grandfather were.

The three Y-111 DNA test confirmed the Y-111 DNA three matches with an overlap of (20.XXX) cM on Chromosome 8.

For this situation, the Y-111 triangulation was used to prove to each of the three third cousins, that they actually had a common great-great grandfather.

It is my opinion, that Y-111 triangulation for three third cousins, with (20.XXX) cM  overlap on Chromosome 8, is valid, highly accurate, and undisputable.

However, the current Wikitree guidelines do not allow this undisputable Y-111 triangulation for three third cousins.

Respectfully, Richard J

Comments

Hi Richard,

I'd suggest posting your own separate G2G post which, if I understand correctly, is about Y-DNA (this G2G question is about autosomal DNA, and is tagged as such).

That will allow you to add appropriate tags to your G2G post, such as y-chromosome and/or y-dna which are tags that may be followed by folks here who can help.

---

I created a new comment at:

https://www.wikitree.com/g2g/1379235/approve-the-triangulation-confirmation-using-three-cousins

with the Tags: 

y-dna,

Y-chromosome,

dna-confirmation,

improvements,

dna,

policy

Thank you for your help!

Richard J