Saturday 18 August 2018

Colour clustering

Last week, one of the genealogy bloggers I follow wrote about a new way of sorting her Ancestry DNA matches. She calls it the 'Color Cluster Method'. It certainly produces colourful results and it's quick and easy to follow, so I thought I'd try it. All going well, it should divide our matches into four clusters, with one to represent each set of great-grandparents.

The first step is to list the second and third cousin DNA matches at Ancestry.com. Unfortunately, however, I only have four such matches in total at this level, barely enough to create a single good cluster, never mind four, so I had to move the goalposts. Though not recommended, I added the fourth cousins Ancestry.com deems high confidence matches to the mix. Some of them are known third cousins anyway and it gave me twenty-six matches to play with.

It sort of worked. I got nine groupings, but among them, four distinct clusters are obvious. And, while not specific to any particular set of great-grandparents, three of my grandparents are definitely separately represented by the first three clusters, as depicted below. 



On my mother's side, the yellow cluster, in column 1, contains known cousins on her paternal Wynne line, all being descendants of my great-great-grandparents John Wynne and Bridget Hynes.  The orange cluster, in column 3, includes two cousins on her maternal Byrne line. They are descendants of my third-great-grandparents, Myles McGrane and Margaret Doyle, but the other matches in this group have not yet been proven. And, on Dad's side, the blue cluster, in column 2, includes five identified cousins on his paternal Byrne line, all descendants of my third-great-grandparents, Andrew Byrne and Anne Clynch.  

So, does the green cluster, in column 4, represent my other grandmother, Dad's maternal O'Neill line?  That's the big question. It would make for an interesting conclusion, as currently we have no known DNA matches on the O'Neill side. 

None of the matches in the green cluster have been identified yet, but as there is no overlap between any of these four clusters, i.e. no match has been allocated more than one colour, the chances are the green group represents a different lineage - maybe the O'Neills. 

The next step is to ascertain surnames common to different members of each group, so as to gather clues to the identity of our as yet undiscovered ancestors. Luckily, four matches in the green cluster have an online family tree, though only two include the surnames of their great-great-grandparents. (Note: if you are following the method as it was intended, and originally included only second and third cousin matches, you need only list the surnames of great-grandparents here). 

Regrettably, there are no surnames in common among any of the matches in the green cluster, bar one. Two matches, including the one with no online tree, share the surname Donvan, an unusual distortion of the Irish name, Donovan. And, Donovan is a known surname in my O'Neill lineage. But, I've already investigated this match, without success; their Donovans seemingly go back to Co. Cork, whereas ours were well established in Dublin city, by the early nineteenth century. 

The remaining five clusters presumably represent earlier generations still. Their lineage may even be indicated by the second colour some matches share with members of a main cluster. 

So, food for thought... 

Perhaps my cousins might like to try out this method? Aileen, Phyllis, Holly - you all have lots of cousins at second and third cousin level. Give it a go! & let me know how you get on. 

See Dana Leeds' easy to follow instructions on The Enthusiastic Genealogist blog at NEW METHOD: DNA Quick Sort and Color Clustering: Identifying Common Surnames and Color Clustering: Working with "4th Cousins" and Color Clustering: Top 25 Fourth Cousins.

10 comments:

  1. I did the same exercise earlier this week. Nearly all my highest matches (including 4C) are paternal. I tried the color clustering for 2C and 3C and got 3 clusters. One for my paternal grandfather and two for my paternal grandmother - one for her father and one for her mother's lines. I started working on first 4C matches but stopped at 20 as it was getting time consuming. I was a bit surprised something so simple would work so well.

    ReplyDelete
    Replies
    1. Absolutely Cathy, I wish I'd known of this method when I first got my results, it's a lovely simple way to start.

      Delete
  2. Rainy weekend expected here so I plan to try Dana's method as well. Thanks for explaining your modification, which I may have to adopt. If I can persuade another cousin to try as well, we might get closer to solving the puzzle!

    ReplyDelete
    Replies
    1. Best of luck, Marian, I hope you make a break-through!

      Delete
  3. Hi, Dara. I love seeing how the method worked for you when using more 4th cousins! I would love to know how many cM each of those matches shared. :) And, please let me know if you figure out the green cluster. It certainly looks like a “real” cluster! Thanks for sharing your results!!

    ReplyDelete
    Replies
    1. Thanks Dana, it's a fun method, as well as being effective. There were 17 matches Ancestry deems 'High' confidence, ranging from 30cM to 43cM, and four 'Very High' matches between 47cM and 51cM, and one 'Extremely High' at 66cM, as well as the 4 '2nd' and '3rd' cousins.

      Delete
    2. Hi again, Dara. As I continue to gain more experience with results from Color Cluster charts, I have a bit more insight. First of all, because you have fewer matches, you are dealing with more distant cousins than I usually work with. Almost 2/3rds share below 45 cM, and I've never worked that low. But, it's good to see it is still working!

      I would collapse column 7 into 1 and column 9 into 3 leaving you with 7, instead of 9, columns. Since you don't have many closer cousins, your cousins aren't being "tied together" as nicely as some. But, I do think you have good possibilities of finding your dad's maternal O'Neil line.

      I hope you get some other matches and are able to determine how these more distant cousins match you and find out if any are from your O'Neil line!

      Delete
    3. Yes Dana, I really need a close O'Neill cousin to take the test :-))))

      Delete
  4. Dara, thanks for sharing this! I tried it out and it's very helpful. Also, I just realized I still wasn't getting Blogger comments. You left a comment on my 52 Ancestor posts about my ancestor that played piano at the silent movies.

    ReplyDelete
    Replies
    1. Thanks Kathryn, I'm glad you found it helpful. After the GDPR bug in Blogger, I had to delete my email address from my settings, save the changes, (in two places, if I remember) then enter my email address again, and save the changes. Then it all worked properly again.

      Delete

I'd love to hear your thoughts on this!