A Segmentology TIDBIT
Triangulated Groups will cover an Ancestral line; Clusters tend to focus on a Common Ancestor.
Think about this for a moment. A Triangulated Group is formed around one segment of your DNA. This segment of DNA was passed down to you from your mother or your father. This segment of DNA was first formed in one of your Ancestors – as a part of their DNA which was passed down through a line of your Ancestors to you. Let’s call this Ancestor your “first” Ancestor, with respect to the DNA represented by the TG segment. Generally: this TG segment probably started as part of a somewhat larger segment of DNA in that “first” Ancestor; it probably got whittled down by recombination along its journey down to you; and portions of the larger segment were also passed down through several children of this “first” Ancestor to other people who became your Matches (because they shared this DNA with you). The bottom line is that you may have a first cousin (1C) who shares part (or all) of this TG segment of DNA with you. You may also have a 3C or a 5C or an 8C who shares part of this TG segment of DNA with you. The point is that within a TG, you may well have Matches who are cousins over a wide range – back to any of your Ancestors between your parent and the “first” Ancestor to have the TG segment DNA. In fact, among your TG Matches there may be cousins beyond your “first” Ancestor – these Match-cousins would share smaller pieces of the TG segment that came from Ancestors of your “first” Ancestor.
Now let’s shift and think about Clusters. Clusters are formed from Matches who are Shared Matches with each other. Each of the Shared Matches in a Cluster *tend* to match most of the other Shared Matches in the Cluster. That’s why we see Cluster diagrams with squares which are almost solid – most Matches match most of the other Matches. This usually happens when the Matches have the same Common Ancestor. Think about the LEEDS method – the focus is on four Clusters, each one representing a different one of your four grandparents. As the lower cM threshold is reduced, more Matches are included in the analysis, and more Clusters are formed. These Clusters *tend* to drift away from grandparents and form around more distant Ancestors. Although it is not a “rule” or “requirement”, it does seem that each cluster is centered on a specific Ancestor. However, sometimes a Match in a Cluster may be related through an Ancestor a generation closer or farther than most of the Matches. This is because the range of relationships is not rigidly tied to cMs – the smaller the cMs in a Cluster, the larger the range of possibilities. This is also due to the fact that a close Match will be included in one of the Clusters – unless the upper cM limit on Clustering is lowered to preclude close cousins. Beyond the 4-generation LEEDS Clusters, the Clusters with smaller and smaller cMs, get more and more “messy” with more and more exceptions to the one Ancestor per Cluster concept. But the “tendency” remains: the Clusters “tend” to form with Matches who have the same Common Ancestor. NB: if you want the Clusters to point to one Common Ancestor, you should either adjust the upper cM limit, or manually cull out Matches who are clearly closer cousins.
A few years ago, I Clustered all of my FTDNA Matches (roughly 8,000 of them). I had already Triangulated them into about 370 TGs. I got about 350 Clusters. In both cases there were about 5% of the Matches who didn’t Cluster or Triangulate – these Matches were all under 15cM (most under 10cM) and were the same Matches in both cases – they were false Matches. There was very close to 100% correlation between the Cluster and the TGs (in other words the Matches in each Cluster had the same TG). My conclusion was/is that the Cluster Common Ancestor was the same as the “first” Ancestor for the TGs [I only wish I knew, for sure, who that CA was…]
Bottom line: you should be able to “Walk the Ancestor Back” with different Matches in a TG; and you should see most of your Matches in a Cluster as cousins with the same Common Ancestor (with maybe a few Matches being a little closer or farther cousins).
[22BE] Segment-ology: Insights on Clustering vs Triangulation TIDBIT by Jim Bartlett 20220413
Regarding triangulation groups and clusters, I also only realized this when implementing tools that focused on both. It’s good to realize that sometimes these shared match clusters represent a single or couple of triangulating segments. Or none at all. Can be everything in between. Here is a slide about this concept (I think it can be explained much better visually –> https://youtu.be/LST8jqY7C2A?t=506).
EJ – Thanks for your insights as well and the link to your presentation – I watched, and recommend, all of your presentations from RootsTech (all free). As you know, I’m a big fan of Clustering – it’s my go-to tool at AncestryDNA. Also I’m having great success with “Walking the Clusters Back” – starting with a Cluster run based on high-cM (close) Matches (for which I know the close Common Ancestors for almost all of them) and then dropping the cM threshold by 5cM increments and intergate/impute the known info into the newer Clusters. I’m down to 20cM threshold now and have been able to link many of the Clusters to TGs.
I still emphasize that Clusters *tend* toward a Common Ancestor – it’s not absolute. But as a grouping tool, Clusters have been very powerfull in finding more CAs hidden in small Trees (even with just a single parent) – it’s work to build those Trees out, but the Cluster is almost like a laser beam pointing to a probable CA, so that is a huge advantage.
Thanks for all the programming you’ve done to make Clustering easier – it’s now up to each of us to use those tools and dig into our Matches’ ancestry to find a CA. Jim
PS: Readers – again, I highly recommend EJ’s short talks on Clustering – see his link.