RootsChat.Com
General => Ancestral Family Tree DNA Testing => Topic started by: ikas on Friday 17 October 25 11:55 BST (UK)
-
Diahan Southard says on her blog that Ancestry are rolling out a custom clustering tool to ProTools account holders and her account has been updated. It is additional to the clustering tool it has already rolled out. Anyone else got it and tried it out?
https://my.yourdnaguide.com/c/diahan-s-corner/ancestry-s-custom-clusters
-
Diahan Southard says on her blog that Ancestry are rolling out a custom clustering tool to ProTools account holders and her account has been updated. It is additional to the clustering tool it has already rolled out. Anyone else got it and tried it out?
https://my.yourdnaguide.com/c/diahan-s-corner/ancestry-s-custom-clusters
She knows her stuff.
Alas the link does not work for me I get “this is a private space”, members only.
It will be available via the additional Subscription to Pro Tools and is expected to be live by the end of October.
Dana Leeds explains the current system here.
https://www.youtube.com/watch?v=Wg2RTuaU5mw
-
Sorry Biggles about the link. I am on her mailing list so I guess that is why it works for me. She has not tried it yet so vid not very informative. Link below is about custom clustering.
https://support.ancestry.co.uk/s/article/Custom-Match-Clusters
-
Clustering and using Ancestry’s Enhanced Shared Matching feature does potentially look very interesting.
The range of 20+cM that is quoted in the Article does cover a host of relationship possibilities.
Hopefully in time this will help me to tie down who my Great Great Grandfather actually was where at present he is a bit of a semi educated guess, we know his name but there are a few possibles and so far inconclusive links with the tools that I have.
-
Yes, I'm hopeful too. Cluster analysis has not helped me in the past and the ancestry clusters look far too clean. I suspect they have dropped a lot of matches to get those sharply defined clusters. I am hoping that being able to select the matches to cluster might be helpful but of course I need to see it in action. I am too busy to try pro tools at the moment but I am planning to have shot early next year.
-
I've got it but it's not making much sense right now. You can create up to 25 custom cluster 'reports' for want of a better description which expire after a month. Each 'report' can have multiple clusters but I'm struggling to make much sense of it right now.
I have one match at 168cM who doesn't make it into standard clusters but has 4 custom clusters with a small unknown match, the custom clusters contain 8,9,8 and 43 matches, the names in the first 3 are just different permutations of the same 9 people.
-
Interesting Glen. I see Gedmatch have also announced a new Tier 1 tool. Launch webinar October 30th. Evert Jan-Blom speaking at the launch. I'll be watching that one.
"Building on the proven AutoSegment clustering method, this new tool doesn't just group your DNA matches—it actually partitions your DNA segments into parental sides AND provides admixture results for each segment on every chromosome."
-
Well, a bit underwhelming.
More complicated to use than the MyHeritage clustering.
The Standard Cluster seems fairly pointless, as it really just gives close relatives.
The Custom Clusters are more powerful, they do let you see how your matches are related to each other, it simplifies what is otherwise a laborious manual process.
I have a large group of related people in Kentucky who seem to link back to a common ancestor in Lincolnshire, but has not generated any blinding revelations yet.
I have manually generated most of this cluster, and it is much bigger 130 or so versus the 20 odds that the Cluster settings allow you to see. The critical linking people between clusters often seem to be in the 10 to 20 cM range as far as I can see, unless you are very lucky.
They say that the clusters stick if they are associated with a group, not sure that is working, especially if you have pre-existing clustering.
The big benefit so far is the fact that it generates a nice picture and an easy to read list of people in a group, also that you can seem to be able to cut and paste the cluster into a word processor or excel. Libre Calc does not seem to work properly for some reason
Warning - note down who you have selected as the seed people for the cluster. You cannot edit the cM settings, or seed people and have to start from scratch each time.
-
I have the pro tools
At present the cluster tool cannot work for me so was hoping the custom would.
Don’t have the custom clustering yet
-
Ancestry have had an outage today and is partly down... haven't been able to view my dna pages since early this morning and hints are not working either at the minute... At least they finally acknowledged they have a problem
-
Well my Enhanced Custom arrived yesterday.
Have to say, totally unimpressed.
-
Custom
One match selected plus four others gives a number of different clusters with 4, 10, 11, 49 and 50 matches in each of the clusters.
This is a partial image of one of the larger clusters.
Considering that many of the DNA Matches that are in the cluster have no method of identifying them and have no tree or a minimal tree this is a practically useless feature.
-
Hi Biggles,
Is there any indication of how many matches have been dropped in the analysis? That is my main gripe with the Jan-Evert Blom clustering methodolgy that Anc seem to have used
-
I'm much more of a newbie at the DNA stuff than some of the other folks here, so I might be approaching the Custom Clusters with lower expectations.
I had already solved my two big DNA mysteries. Using this tool has been a way of visually confirming my findings, rather than helping me to solve them.
But I'm also trying to help a friend with some of her DNA mysteries and had been attempting to use the Leeds Method in an Excel spreadsheet for some of her DNA matches. The Custom Clusters showed me that two separate colour groups actually belonged with a third (her biggest and strongest group of matches). (This might have been obvious if most matches had had trees, or big enough trees.)
Anyway, being able to fold those colour groups into one has shown me the two other distinct groups I should focus on. So it's been helpful in that regard.
For visualizing my own DNA matches, I start with the widest range of cMs: 20-1300, then do 50-1300. For me, seeing the results has been a confirmation that I'm on the right track with my DNA sleuthing. I'm only using the one primary, targeted match in any given cluster. I started by adding others but it wasn't helpful for my purposes.
-
Custom
One match selected plus four others gives a number of different clusters with 4, 10, 11, 49 and 50 matches in each of the clusters.
This is a partial image of one of the larger clusters.
Considering that many of the DNA Matches that are in the cluster have no method of identifying them and have no tree or a minimal tree this is a practically useless feature.
I’m not impressed at all with the clusters tool for my names. very basic and not telling me anything that I don’t already know on my immediate family… my chart doesn’t look as impressive as your one.
-
I have had a further play around and thisnis what we have:-
We are trying to identify if an unknown Father’s has a descendent who has taken a DNA test.
This is what I have to work with, Grace gave birth to Bella out of wedlock, and Bella’s father has remained unknown.
Grace married and her great Granddaughter Babs took a DNA test, the person Grace married was only 9 at the time she conceived hence we can rule him or any of his family as he was the oldest of his siblings.
Bella married and three of her great grandchildren each took a DNA test an they are full Cousins to each other and Dave the son of a deceased Cousin has also taken a DNA test and he is a 1C1R to the other three. Babs is a 1/2 2C1R with Jim, Dennis and Jen.
This is the amount of DNA each share with Babs:-
Jim shares 47cM with Babs
Dennis shares 29cM with Babs
Jen shares 0 cM with Babs
Dave shares 0cM with Babs
Yet
Jim and Babs only share 5 matches and apart from Dennis the other 4 all share less than 20cM so cannot be selected in the custom cluster.
The image is the bottom right of a 59x59 cluster and the Shared Matches between Jim and Babs is the small cluster very bottom right is the group who share most with Babs.
There is one potential DNA match who might be of the Family we are looking for but they return the “less than 20cM” message that they are not considered.
There is no match to anyone that looks like it will help.
The wait continues
-
Biggles,
You've been hugely helpful for me recently. So, sorry for another question...
What indicates that that snip in your last post is part of one big cluster?
I have a whole bunch of small (3x3 / 4x 4) clusters which I am treating a separate tiny entities. Which I am hoping will at some point in the future join up (although it is slow going at the moment, with so many matches having no tree at all / only a single person "private" tree / are in the US, Aus, Can, NZ so I can't check trees, I am having to resort to working from unchecked Anc trees to try to triangulate to a common ancestor).
-
My last three posts where I have included a screenshot of the clustering are each totally unrelated to each other.
The first image is from my Paternal DNA.
There should be a much larger cluster or clusters to work through. Some sub 20cM people who are in the family tree and who could potentially have a bearing on research have not been included.
Second is from my Wife’s matches.
Inconclusive, it is a bit of an oddball. Two 2C’s who are related to my Wife and to her Cousins Grandson do not show as matches to each other yet they should. I had hoped that this cluster would help, but not so far.
The third is from a non biologically related family member where I am trying to find the unknown Father.
Initially inconclusive, however there may be a clue somewhere to help unlock the mystery, it will just take work, or I just wait until the tools get better.
-
Biggles,
You've been hugely helpful for me recently. So, sorry for another question...
What indicates that that snip in your last post is part of one big cluster?
I have a whole bunch of small (3x3 / 4x 4) clusters which I am treating a separate tiny entities. Which I am hoping will at some point in the future join up (although it is slow going at the moment, with so many matches having no tree at all / only a single person "private" tree / are in the US, Aus, Can, NZ so I can't check trees, I am having to resort to working from unchecked Anc trees to try to triangulate to a common ancestor).
Thanks for your kind comment.
I cannot display the whole cluster on my iPad just a portion of it so snip is a misnomer.
I have not yet got my head around all the multiple clusters that Ancestry produces. Guess I will have to print them out as I do not have the screen real estate to display the cluster onscreen together and still being able to read the names.
-
Ah Ok.
As you've been such a oracle and guiding light of knowledge for me on my steep learning curve so far, I thought there was a sign that I was missing that grouped these seemingly independent mini (c.3x3) clusters (for me in my Excel equivalent of this) in to bigger groups.
-
The Standard cluster has been absolutely useless to me- it keep flitting between four matches related to each other and two sets of four related to each other. I know I've got more matches than that!
I've just had a tinker with the custom groups and clearly I'm not understanding it either. I wanted to get an easy representation of different descendants of my illegitimate great great grandfather vs the shared matches I think are his father's side of the family. I chose "sidekick" matches (as Anc puts it) from that suspect family and instead of showing them, or the shared matches I know we have, it kept padding out the clusters with his wife's side of the family >:( Argh!
-
I am getting frustrated with how many shared (maternal/paternal or unassigned) matches I am getting and Ancestry removing confirm your match beyond 4th cousin… I have so many I can confirm which are 4th x1r or 2r which I have to mark up as distant on Anc..
I have managed to verify my lines back to beyond 4th grandparents which is good but my distant and wider matches are a right mixed bunch! ok if they come from the same area, I accept there’s a possibly of meeting and getting friendly but not when I get one verified paternal match who never left the UK and the rest are maternal coming from the US 🤣 Trying to tell myself they are false positives.
-
The Standard cluster has been absolutely useless to me- it keep flitting between four matches related to each other and two sets of four related to each other. I know I've got more matches than that!
I've just had a tinker with the custom groups and clearly I'm not understanding it either. I wanted to get an easy representation of different descendants of my illegitimate great great grandfather vs the shared matches I think are his father's side of the family. I chose "sidekick" matches (as Anc puts it) from that suspect family and instead of showing them, or the shared matches I know we have, it kept padding out the clusters with his wife's side of the family >:( Argh!
Apologies if this is a stupid question, but I'm not sure I understand what you're saying.
If I understand you correctly, you have an illegitimate great-great-grandfather (let's call him Mr. X). He had a wife. You've identified DNA matches with their descendants and with the descendants of other members of their wider families.
You want the Custom Clusters to distinguish between the descendants of the father of Mr. X (the descendants of Mr. X's siblings) from Mr. X's descendants. But you're finding that the Custom Clusters can't do this.
Is that because the amount of shared cMs is too small for the system to easily distinguish between Mr. X's descendants and his sibling's descendants? Is it because the system's cut-off of 20 cM as the lowest shared amount is too high for your purposes?
What would the solution be? A different tool? A modification to the Custom Clusters tool? A different type of DNA test altogether?
I hope you don't mind my asking. I'm trying to learn what I can about the DNA aspect of genealogy research and sometimes it's a struggle to wrap my mind around these things. I find that I tend to use various programs in limited ways and don't always know about everything they can do for me.
-
I am getting frustrated with how many shared (maternal/paternal or unassigned) matches I am getting and Ancestry removing confirm your match beyond 4th cousin… I have so many I can confirm which are 4th x1r or 2r which I have to mark up as distant on Anc..
I have managed to verify my lines back to beyond 4th grandparents which is good but my distant and wider matches are a right mixed bunch! ok if they come from the same area, I accept there’s a possibly of meeting and getting friendly but not when I get one verified paternal match who never left the UK and the rest are maternal coming from the US 🤣 Trying to tell myself they are false positives.
We all are all getting to grips with the new feature and learning its quirks so we can probably not provide anything definitive to help.
The present initial iteration of the Custom Clustering feature is limited in the range of cM that can be selected for the available Matches.
Maybe this will change in future revisions, who knows but time will tell.
In your case the relationship with the Matches is probably distant and maybe a half relation and the cM with them or with those Matches who may hold a key to unlocking who is who and hence will not necessarily show in the cluster directly.
If you can identify those Matches of interest with Shared matches who share over 20cM with you and include them in the Custom Cluster selection process, this may be a workaround for you.
For info for other Forum Members.
In Pro Tools we can look at all Shared Matches irrespective of the cM that you share with them.
Warts and all and a potential problem may be false positives, I do not know if this is actually likely ir otherwise but it is probably good to be cautious.
In getting the system to function we select a reference match and up to 4 shared matches, the caveat being that the shared cM has to be between 20 & 1300 cM.
Hope this helps.
-
Apologies if this is a stupid question, but I'm not sure I understand what you're saying.
No apologies needed, no stupid question :)
If I understand you correctly, you have an illegitimate great-great-grandfather (let's call him Mr. X). He had a wife. You've identified DNA matches with their descendants and with the descendants of other members of their wider families.
Correct. I've got DNA descendants from three children of the marriage. Mrs X had a big family so there are a lot of DNA cousins on her side. That was at least helpful in separating out shared matches of the descendants that didn't come from Mrs X's side of the family, meaning they should be Mr X. I have Mr X's mother's side back several generations so when I got a lot of matches to one family that wasn't hers, it seemed good odds that it is the unknown father's family.
You want the Custom Clusters to distinguish between the descendants of the father of Mr. X (the descendants of Mr. X's siblings) from Mr. X's descendants. But you're finding that the Custom Clusters can't do this.
I hoped that I could narrow the cluster down to only Mr X's father's family, showing cM amounts and positive connections between different members of that family and one or more descendants of Mr X. Instead, it ignored the names I asked it to compare between (and I know there's a DNA match, not least because it only allows 'sidekick' matches that are connected) and gave me large clusters only of Mrs X's family.
Is that because the amount of shared cMs is too small for the system to easily distinguish between Mr. X's descendants and his sibling's descendants? Is it because the system's cut-off of 20 cM as the lowest shared amount is too high for your purposes?
I wondered that, but I don't think so. The cM amounts are similar to Mrs X's family so at least a few of them should have shown up. One of the descendants is very biased towards Mrs X's DNA so I could understand it in his case, but another has more DNA in common with Mr X's side. After some fiddling with it I did eventually get some of them to show up but perhaps more would with a lower cM threshold as you say. I imagine a lot of it is me not understanding how the function works. It might also be comparing between my mother's DNA results and the home person (another descendant) I selected so it's giving all results for people we are both related to, although that still doesn't explain the bias towards Mrs X's results. I might have another tinker with it.
I hope you don't mind my asking. I'm trying to learn what I can about the DNA aspect of genealogy research and sometimes it's a struggle to wrap my mind around these things. I find that I tend to use various programs in limited ways and don't always know about everything they can do for me.
Not at all, although I think this entire thread is other people also being confused! Asking questions is great :)
-
Thank you so much, Ayashi.
It's funny, because later in the day I tried to do a Custom Cluster, only to discover that all of the people who should have been in it (maybe a handful or two) shared below 20 cMs with me. Whoops!
I wonder if one of the reasons for the cut-off is that the resulting clusters would be so huge they'd be unmanageable, both for the user and/or for the system that would be generating them? I already have some clusters with 72 people that I have to shrink in order to see them all at once, but then it's really hard to see all the tiny names, etc.
I've been using the Shared Matches tool and it's been really helpful. It would just be nice if a machine would do all that work for me (LOL).
-
I am really unclear what issue this clustering will help with.
I tried a match of 55cmorgans, who shares my grandmothers maiden name
I doesn’t really help much
The standard match won’t work at all for me
-
I have had another session with Custom Clusters.
In post #14 there is the detail of the brickwall that I am working on.
This time I selected just two 1C’s and one 1C1R for the sequence.
It generated 8, 22, 28 & 57 cluster grids.
Nothing looks to help, in fact the opposite.
The vast majority are wholly Paternal but via the Paternal Grandmother when what was really wanted is an OMIT feature so that I could tell the system to ignore those matches that have specific Group tags. Plus an Include feature so that the DNA matches that I have added a specific Group such as Work In Progress group can be selected which in turn will force the algorithm to look at those who are Tagged.
Then, Hopefully the Clusters would be based on Paternal Paternal matches and hopeful unknowns who share with Paternal Paternal matches.
At lease this is where my head is at present.
-
Hi, I think you need to take the same approach as you would manually. Look at the people who are in your close cluster, who you cannot assign to paternal maternal line and then pick people who are at the once removed or further, and try clustering with them.
I am not sure if you are lucky or unlucky, the biggest cluster I have managed is about 20. All my clusters are in the 40 to 15 cM range so this is not really helping at the moment.
-
I don't think it's unfair to say that it's another DNA 'feature' that many think is the answer to everything when it's nothing of the sort. With a large and robust set of matches it may offer a bit of time saving but I still feel that it provides no more answers than diligent grouping and tree building in most cases.
Slightly off topic but the most beneficial thing I feel we could have is seeing which side we link from the perspective of our matches. It's not as if Ancestry don't know, they just won't reveal it.
-
I don't think it's unfair to say that it's another DNA 'feature' that many think is the answer to everything when it's nothing of the sort. With a large and robust set of matches it may offer a bit of time saving but I still feel that it provides no more answers than diligent grouping and tree building in most cases.
I completely agree!!! Correct me if I'm wrong but this feature does not provide us with any NEW information about Matches or anything for that matter - it just slices and dices information already available and presents it in a new way. Like I say, please correct me if I am missing something
-
This isn't about the function, exactly, but I wish it was possible to copy and paste the relationship estimates from the screen when I'm looking at the Custom Cluster results.
As it is, I have to hover my mouse over the little box to get the pop-up text, try to memorize something like, "1/2 3rd cousin 1x removed or 4th cousin," click my mouse in another window or slide it over to my other monitor and click that screen to get to my database... and by that point, I can only be certain that I've remembered the first part correctly. So I have to do it all over again, in reverse, to get back to the Ancestry screen and hover over the little box, then back to my other screen, to type in the rest of it... Multiple times over for shared matches with one person. I find that to be a bit onerous.