A Segment-ology TIDBIT
Are Clusters based on Common Ancestors (CAs) or Triangulated Groups (TGs)? I said CAs, and Jonathan Brecher said TGs. I now think Jonathan has the best answer. My apologies for doubting his conclusions.
My point was that with a few, large, close, Clusters, each Cluster must be formed on a CA, and include many TGs. A Match Clustering which results in 4 or 8 or 16 Clusters (which they tend to do) are clearly formed on 4 grandparent, 8 Great grandparents or 16 2xGreat grandparents – and this is born out with the Leeds Method and other experience. These large Clusters must each include many TGs – and my experience bears this out.
However… There’s often an “however.” As the Clustering thresholds are decreased, the number of Clusters formed are increased. In my recent example I had 156 Clusters using a 20cM threshold. Note that one fourth of my Ancestry is from 1850’s immigrants, with very few Matches (and all of them were close cousins in one Cluster). I should have had about 208 Clusters. And 161 Matches did not cluster. This gets us pretty close to 256 6xGreat Grandparents 8 generations back. And a number of my 156 Clusters appear to link with only one TG [Note this is AncestryDNA data, most of it without TGs].
I am now reviewing my AncestryDNA Matches and trying to assign Cluster IDs to each one, by looking at the info I have in the Notes box for each Match and reviewing all their Shared Matches (and their Notes – easily viewed with MEDBetterDNA). In most cases, where the Match shares a single segment with me, I’m tending to identify a single Cluster. And when I have TG information, it’s tending to be one TG.
So, I’m going to eat some crow, and apologize to Jonathan Brecher. I now think he was on the right track, and that we should try to link Clusters to TGs (specific DNA segments). After all, each TG is from a specific Ancestral line. Of course, at AncestryDNA (without segment data), we’d still Cluster mainly on CAs. However, with a comparison between AncestryDNA Clusters and Clusters with other companies (with segments and TGs), we should be able to find a correlation between our AncestryDNA CA Clusters and TGs. Through this correlation, we could “impute” TGs to AncestryDNA Clusters.
So – thank you, Jonathan Brecher – for Clustering for several of us, for your comprehensive analysis and for your insight!
If anyone has been Clustering around CAs, that is still OK. Think of your Cluster CAs as potentially having multiple TGs – particularly the closer CAs (4C-6C range). And as you run Clustering with smaller thresholds, and find more Clusters, you’ll find your former Clusters subdividing into smaller Clusters, which smaller Clusters would tend to match up with one TG – Walking the Clusters Back!
[22AE] Segment-ology: Clusters Link to TGs and an Apology TIDBIT by Jim Bartlett 20190222