PDA

View Full Version : New feature Theory of Family Relativity



timberwolf
02-28-2019, 05:17 AM
Just reading over on 23andme that myheritage has released a new feature 'Theory of Family relativity, some people seem to have it. I don't as of yet.

Foes anyone here have it?

Kathlingram
02-28-2019, 02:48 PM
Just reading over on 23andme that myheritage has released a new feature 'Theory of Family relativity, some people seem to have it. I don't as of yet.

Foes anyone here have it?

I do have it.. I have NOT gotten back their interpretation of my Family Relatives.. They collate and send it to your email.. What I am hearing is that they us 2 Family Search.org trees and a DNA match as 90% peoof ( Maybe I have that backwards.. Dunno)

Kathlingram
02-28-2019, 02:51 PM
It is here.. and of course because to Roots Tech
29100

timberwolf
02-28-2019, 06:26 PM
It is here.. and of course because to Roots Tech
29100

Do you need a subscription?

timberwolf
02-28-2019, 06:44 PM
To answer my own question. It is a one time fee of $29 with a uploaded kit to unlock these features the auto clustering tool and theory of family relatively which are separate tools.

Otherwise with your myheritage kit your need a subscription which is expensive.

Pylsteen
02-28-2019, 07:05 PM
I have it. They have a theory for 5 of the few thousand matches, so not a lot (yet?). Their proposed theories are correct (all shared early 19th century ancestors), but I already figured the relationships out myself, and I trust myself more than the tool in doing so. It may become more valuable in the next decade, when people will have many more matches and no time left to figure them all out.

Kathlingram
02-28-2019, 07:11 PM
Well they sent it to me but I cannot see how to open or read it..
"
Dear Kathleen,‎
A zip file with your report of AutoClusters is attached. In it, you will find three files — an HTML file which contains a visual representation of the AutoCluster analysis, a CSV file which contains a spreadsheet version of the AutoCluster analysis, and a ReadMe PDF file.

Some files may be displayed incorrectly when opened directly from this email. For best results, view the files on a desktop device. To do this, save the zip file to your hard drive, extract the contents, and then open the files.
Please note that this file is intended for your personal use only. The attached file is no longer protected by MyHeritage security and privacy mechanisms, so please take additional measures to secure your data and avoid sharing it.

Best regards,
The MyHeritage team "

timberwolf
02-28-2019, 07:14 PM
I have just had to reload my ancestry kit which takes 5-7 days. I guess I will be asked to pay after that?

Seems a bit strange to me that an uploaded kit you can lock it for $29 or free if you had previously done so.

With myheritage kits it seems you can get some of the information, but some is hidden and you would need to pay the rather expensive subscription

Kathlingram
02-28-2019, 07:22 PM
I have just had to reload my ancestry kit which takes 5-7 days. I guess I will be asked to pay after that?

Seems a bit strange to me that an uploaded kit you can lock it for $29 or free if you had previously done so.

With myheritage kits it seems you can get some of the information, but some is hidden and you would need to pay the rather expensive subscription

Not it your Tree is smaller that 250 people.. Mine is an older Tree

Pylsteen
02-28-2019, 07:27 PM
Oh, wow, I now see that they have an option for AutoClustering, let's see what that will do.

Loderingo
02-28-2019, 07:32 PM
I have auto clustering but no theories on any of my kits. Could this be as my trees are too big?

timberwolf
02-28-2019, 07:37 PM
Not it your Tree is smaller that 250 people.. Mine is an older Tree

Thanks still it is a minimum of one year which is $120NZ as opposed to a one off $29US about $40 NZ

And it seems that people who uploaded Kits can use the autoClustering tool while I cannot.

Loderingo
02-28-2019, 07:37 PM
I have found an answer to my question on the Legal Genealogist blog:

Only kits that tested or uploaded before November 2018 have Theory of Family Relativity entries. Others coming online later will have to wait for the database to be re-run against the records.

timberwolf
02-28-2019, 07:49 PM
I have found an answer to my question on the Legal Genealogist blog:

Only kits that tested or uploaded before November 2018 have Theory of Family Relativity entries. Others coming online later will have to wait for the database to be re-run against the records.

Thanks

But is still strange that for uploaded kits this feature will available in its entirely while for a Myheritage kit you have to pay a subscription to unlock most of its features.

Kathlingram
02-28-2019, 10:26 PM
I have 7 "Theories" on the MyHeritage website but the ZIP File they sent me shows more matches in each theory.. However some are 'low confidence" and some have no tree


For example: Path #1 This path is based on one community tree and 2 MyHeritage family trees, with 77% confidence and Path #2 This path is based on one community tree, one record collection and 2 MyHeritage family trees, with 30% confidence

coffeeprince
03-01-2019, 12:22 AM
I have found an answer to my question on the Legal Genealogist blog:

Only kits that tested or uploaded before November 2018 have Theory of Family Relativity entries. Others coming online later will have to wait for the database to be re-run against the records.

Not sure if that's true. I uploaded in June 2017 and don't have the Theory of Family Relativity option. I do have the AutoClustering option though.

timberwolf
03-01-2019, 12:34 AM
Not sure if that's true. I uploaded in June 2017 and don't have the Theory of Family Relativity option. I do have the AutoClustering option though.

Are you able to use the AutoClustering option without needing a subscription?

vettor
03-01-2019, 05:29 AM
Are you able to use the AutoClustering option without needing a subscription?

its a one off fee...IIRC $29

took 13 hours to do mine

AutoClusters is a new technology on MyHeritage that organizes your DNA Matches
into shared match clusters, that likely descended from common ancestors. Each of
the colored cells in the chart represents an intersection between two of your matches,
meaning that both individuals match each other (in addition to matching you). These
cells are grouped together physically and by color to create a powerful visual chart of
your shared match clusters.
Each color represents one shared match cluster. Members of a cluster match you
and most or all of the other cluster members. Everyone in a cluster will likely be on
the same ancestral line, although the most recent common ancestor between any of
the matches and between you and any match may vary. The generational level of the
clusters may vary as well. One may be your paternal grandmother’s branch, and
another may be your paternal great-grandfather’s branch.

Your AutoCluster analysis was generated using thresholds of 10 cM (minimum) and
350 cM (maximum). In addition, DNA Matches were required to share at least 10 cM
with one another in order to be indicated with a colored or gray cell. A total number of
100 DNA Matches ended up in 25 clusters in the final analysis.


etc etc


its usefulness is in basically joining people who come from the same group in regards to Chr
they discard all the single matches who do not fit with others

timberwolf
03-01-2019, 05:51 AM
its a one off fee...IIRC $29

took 13 hours to do mine

AutoClusters is a new technology on MyHeritage that organizes your DNA Matches
into shared match clusters, that likely descended from common ancestors. Each of
the colored cells in the chart represents an intersection between two of your matches,
meaning that both individuals match each other (in addition to matching you). These
cells are grouped together physically and by color to create a powerful visual chart of
your shared match clusters.
Each color represents one shared match cluster. Members of a cluster match you
and most or all of the other cluster members. Everyone in a cluster will likely be on
the same ancestral line, although the most recent common ancestor between any of
the matches and between you and any match may vary. The generational level of the
clusters may vary as well. One may be your paternal grandmother’s branch, and
another may be your paternal great-grandfather’s branch.

Your AutoCluster analysis was generated using thresholds of 10 cM (minimum) and
350 cM (maximum). In addition, DNA Matches were required to share at least 10 cM
with one another in order to be indicated with a colored or gray cell. A total number of
100 DNA Matches ended up in 25 clusters in the final analysis.


etc etc


its usefulness is in basically joining people who come from the same group in regards to Chr
they discard all the single matches who do not fit with others

Thanks

Unfortunately it just takes me to a subscription page. I know if I upload a kit then it is $29

JFWinstone
03-01-2019, 08:48 AM
We have auto clustering but not theories.

firemonkey
03-01-2019, 09:13 AM
I have auto clustering but not theories. There's just 108 dna matches in auto clustering ( 15cMs or over) .

03-01-2019, 09:26 AM
Relativity works, by using people’s family tree’s I think, good idea, and correlating “cm”, matching. Have Clicked to Generate auto clustering, for me at least seems free.

JFWinstone
03-01-2019, 09:29 AM
It looks like mum has 2 matches with theories, both descendants of the same couple. One of them is a match on Ancestry and we knew the common ancestors for that person. Dad and I don't have any.

firemonkey
03-01-2019, 01:45 PM
My father and I have 2 theories of relativity. Both are 20% . Both are total bunkum.

Loderingo
03-01-2019, 02:18 PM
I have auto clustering but not theories. There's just 108 dna matches in auto clustering ( 15cMs or over) .

It looks like they change the thresholds for different people.

For my father's autocluster the read me says:

Your AutoCluster analysis was generated using thresholds of 25 cM (minimum) and 350 cM (maximum). In addition, DNA Matches were required to share at least 25 cM with one another in order to be indicated with a colored or gray cell. A total number of 97 DNA Matches ended up in 30 clusters in the final analysis.

However, for my mother's autocluster the read me says:

Your AutoCluster analysis was generated using thresholds of 25 cM (minimum) and 350 cM (maximum). In addition, DNA Matches were required to share at least 20 cM with one another in order to be indicated with a colored or gray cell. A total number of 76 DNA Matches ended up in 15 clusters in the final analysis.

And for my autocluster the readme says:

Your AutoCluster analysis was generated using thresholds of 25 cM (minimum) and 350 cM (maximum). In addition, DNA Matches were required to share at least 10 cM with one another in order to be indicated with a colored or gray cell. A total number of 90 DNA Matches ended up in 24 clusters in the final analysis.

So it seems like they change the settings to try and get around 50-100 people in the final analysis

miremont
03-01-2019, 03:03 PM
I have both new features, but the auto clustering has not completed yet after approximately 24 hours.

FionnSneachta
03-01-2019, 05:02 PM
There are 3 theories of relativity for me.
1. My dad's second cousin with the correct predicted relationship with 88% confidence.
2. My dad's third cousin correctly predicted with 80% confidence.
3. My 5th cousin correctly predicted with 97% confidence (also on my dad's side). I knew the match but didn't ever fully look into the connection.
The theories are just based on family trees.

My dad has 10 theories.
1. The first is his second cousin again (88%).
2. A 4th cousin of my dad (92% confidence). Another match where I had an idea of the connection but never confirmed it.
3. The third is the 5th cousin of me above (97%).
4. A 4th cousin once removed of my dad who is the daughter of 2. (92%).
5. My dad's third cousin in my list at 2. (80%).
6. A 4th cousin with 52% confidence. I didn't have her in my tree before this.
7. This is the first one that doesn't seem correct with 25% confidence. It has a person born in Boyle in a family tree as being the same woman born in Cork in 1911. Her father born in Cork is recorded as the same person born in Mayo. His grandparents happen to have the same names as my 3x great grandparents but are from Bohola in Mayo. I don't know how I connect to this match. They're a 19.9 cM match.
8. This one is also incorrect with 25% confidence. It has my 2x great grandmother appearing as the same woman in another tree despite having different surnames and one born in Roscommon and the other in Kerry. They're a 18.8 cM match.
9. This is also wrong with 30% confidence. This match is made to connect to the same couple as 7. The problem MyHeritage connect his Michael as the same person as Michael Patrick. However, in the case of 7. they made Michael Patrick the same person as a person named Patrick. He's a 8.1 cM match.
10. This one is also wrong. It has 20% confidence. It claims 2x great grandmother is the same woman born to a man born in Jefferson County, Georgia. They're a 13 cM match. I don't know where the connection is but we do have a common surname of Fleming but their family were from Antrim in 1750 and with names like Samuel I don't think that's the connection.
The first 6 were right and I did find new connections but I think I also got a bit of a headache trying to figure how they were linking up.

My mum has no theories. She only has two lines that had lots of descendants going back to the late 1700s.

My great aunt has 4 theories. She gets 1., 6., 2. in my dad's list.
She also gets an extra match with 20% confidence. They actually give me a couple as predicted ancestors who I thought could have been my ancestors before. They're from the neighbouring townland where my 3x great grandfather was from. The couple also had a son with the same name as my ancestor so I thought that it could have been him. I think that the reason they're a predicted match is because a lady trying to figure her biological father has added loads of trees together and that couple were in her tree. I told her why I no longer think that they're my ancestors but were never removed. It didn't really bother me because the tree on Ancestry is private but it seems she has another tree on MyHeritage that doesn't seem to be private. I didn't realise that I matched a descendant of this couple on MyHeritage. I match another one on Ancestry. However, they got a descendant to do a Y-DNA test and we have a common ancestor around 1226 AD but no more recent than that unless the descendant they got to test wasn't actually a descendant of the couple. I think it is just a coincidence that there are common names, living around the one area and are autosomal DNA matches though.

I've requested the auto-clustering so that should be interesting.

vettor
03-01-2019, 05:29 PM
my auto clustering is at 10cm ....total 100 matches
first cousin also at 10cm......she has 34 matches in group 1 .......total 80 matches
and my wife at 15cm ....total 96 matches

Loderingo
03-01-2019, 06:40 PM
There are 3 theories of relativity for me.
1. My dad's second cousin with the correct predicted relationship with 88% confidence.
2. My dad's third cousin correctly predicted with 80% confidence.
3. My 5th cousin correctly predicted with 97% confidence (also on my dad's side). I knew the match but didn't ever fully look into the connection.
The theories are just based on family trees.

My dad has 10 theories.
1. The first is his second cousin again (88%).
2. A 4th cousin of my dad (92% confidence). Another match where I had an idea of the connection but never confirmed it.
3. The third is the 5th cousin of me above (97%).
4. A 4th cousin once removed of my dad who is the daughter of 2. (92%).
5. My dad's third cousin in my list at 2. (80%).
6. A 4th cousin with 52% confidence. I didn't have her in my tree before this.
7. This is the first one that doesn't seem correct with 25% confidence. It has a person born in Boyle in a family tree as being the same woman born in Cork in 1911. Her father born in Cork is recorded as the same person born in Mayo. His grandparents happen to have the same names as my 3x great grandparents but are from Bohola in Mayo. I don't know how I connect to this match. They're a 19.9 cM match.
8. This one is also incorrect with 25% confidence. It has my 2x great grandmother appearing as the same woman in another tree despite having different surnames and one born in Roscommon and the other in Kerry. They're a 18.8 cM match.
9. This is also wrong with 30% confidence. This match is made to connect to the same couple as 7. The problem MyHeritage connect his Michael as the same person as Michael Patrick. However, in the case of 7. they made Michael Patrick the same person as a person named Patrick. He's a 8.1 cM match.
10. This one is also wrong. It has 20% confidence. It claims 2x great grandmother is the same woman born to a man born in Jefferson County, Georgia. They're a 13 cM match. I don't know where the connection is but we do have a common surname of Fleming but their family were from Antrim in 1750 and with names like Samuel I don't think that's the connection.
The first 6 were right and I did find new connections but I think I also got a bit of a headache trying to figure how they were linking up.

My mum has no theories. She only has two lines that had lots of descendants going back to the late 1700s.

My great aunt has 4 theories. She gets 1., 6., 2. in my dad's list.
She also gets an extra match with 20% confidence. They actually give me a couple as predicted ancestors who I thought could have been my ancestors before. They're from the neighbouring townland where my 3x great grandfather was from. The couple also had a son with the same name as my ancestor so I thought that it could have been him. I think that the reason they're a predicted match is because a lady trying to figure her biological father has added loads of trees together and that couple were in her tree. I told her why I no longer think that they're my ancestors but were never removed. It didn't really bother me because the tree on Ancestry is private but it seems she has another tree on MyHeritage that doesn't seem to be private. I didn't realise that I matched a descendant of this couple on MyHeritage. I match another one on Ancestry. However, they got a descendant to do a Y-DNA test and we have a common ancestor around 1226 AD but no more recent than that unless the descendant they got to test wasn't actually a descendant of the couple. I think it is just a coincidence that there are common names, living around the one area and are autosomal DNA matches though.

I've requested the auto-clustering so that should be interesting.

It sounds like it's working reasonably well then

Jatt1
03-01-2019, 06:46 PM
I have auto clustering but not theories. There's just 108 dna matches in auto clustering ( 15cMs or over) .

And this 15cMs or over is the total or the minimum segment size?

firemonkey
03-01-2019, 07:13 PM
Minimum segment size.

FionnSneachta
03-01-2019, 07:48 PM
It sounds like it's working reasonably well then

Yeah and at least they do give confidence percentages so if it's given as 20%, you can expect that it may not be correct.

timberwolf
03-01-2019, 08:11 PM
I was able to get autoclustering by re uploading an ancestry test for $39NZ, yet with the myheritage kit I would only get that feature by taking out a subscription the minimum cost would have been $120NZ. Seems a bit about face to me.

At least with the reload the Nigerian disappeared from my results.

coffeeprince
03-02-2019, 01:18 AM
Are you able to use the AutoClustering option without needing a subscription?

Yes, I can.

Phoebe Watts
03-02-2019, 12:51 PM
And this 15cMs or over is the total or the minimum segment size?

Total.

Both my reports were generated with a threshold of 30cM min. Then for my upload there needed to be a 10cM for selection. 110 of those matches ended up in 24 clusters. Another 47 matches didn’t make it into clusters.

My relative’s upload needed a 20cM match to be selected. 100 of those matches ended up as 25 clusters. 105 further close matches did not make it into clusters.

FionnSneachta
03-02-2019, 05:01 PM
For autoclustering, all clusters for my family were generated from 25 cM to 350 cM matches.
For me, 101 matches ended up in 27 clusters. 11 matches were excluded. The largest cluster has 7 members (I don't know the connection to any of them).
My mum had 100 matches in 27 clusters with 5 excluded. The largest cluster has 9 members. I don't know the connection to any of them.
My dad had 86 matches in 26 clusters with 8 excluded. The largest cluster has 5 members. I do know the common ancestors of these matches which are my 4x great grandparents
My great aunt had 101 matches in 30 clusters with 14 excluded. The largest cluster has 7 members. I don't know the connection there either.

Some look like they make sense. However, I don't know how I'm connected to most of my matches so I'd need to spend time looking at the matches to determine what ancestors the clusters are associated with.

It's good to see that they're developing new features though that might help.

FionnSneachta
03-02-2019, 05:02 PM
Duplicate

kiterunner
03-02-2019, 07:16 PM
My AutoCluster used thresholds of 35cM min and 350cM max and had to share at least 10cM with each other. I have 9-12% Ashkenazi Jewish DNA (depending on which site you believe) and of my 16 clusters, it looks as though 14 of them are Jewish, with a lot of grey cells all over the place (showing that all those clusters are connected to each other), except on the lines of the other 2 clusters. I understand that MyHeritage doesn't filter out the common Jewish DNA like ancestry does, and I doubt that the people in the 14 Jewish clusters are particularly closely related to me or to each other. The other 2 clusters look more hopeful.

JFWinstone
03-02-2019, 10:07 PM
I have 69 matches from 12 clusters. Cluster 1 consisted of 26 matches that look to be all English plus two more smaller clusters, 3 Swedish clusters, 3 Dutch, 2 French, 1 German
Mum has 100 matches from 19 clusters and they all look like they are through her English side
Dad has 134 matches from 33 clusters a mixture of English, Dutch, German, Swedish, Danish, French and 1 Eastern European cluster

timberwolf
03-02-2019, 10:13 PM
I have 69 matches from 12 clusters. Cluster 1 consisted of 26 matches that look to be all English plus two more smaller clusters, 3 Swedish clusters, 3 Dutch, 2 French, 1 German

How long did it take for you to get your report?

JFWinstone
03-02-2019, 10:19 PM
How long did it take for you to get your report?

About 2 days I think

msmarjoribanks
03-02-2019, 10:34 PM
I've been working with clusters from Genetic Affairs, which are helpful although you definitely need to analyze them.

https://blog.kittycooper.com/2018/12/automatic-clustering-from-genetic-affairs/

This seems identical but for My Heritage, and I am hoping it will add to the analysis.

timberwolf
03-02-2019, 10:38 PM
I've been working with clusters from Genetic Affairs, which are helpful although you definitely need to analyze them.

https://blog.kittycooper.com/2018/12/automatic-clustering-from-genetic-affairs/

This seems identical but for My Heritage, and I am hoping it will add to the analysis.

Interesting. How long have you been waiting?

Dibran
03-03-2019, 10:02 AM
How does one decipher the report? I have 10 clusters. Total 100 matches. Cluster one is 50/60 or so matches. Most are all Albanians. The latter clusters had a couple Greeks, Armenians, Turks and a Macedonians(assuming torbesh by the name).

msmarjoribanks
03-03-2019, 02:42 PM
Interesting. How long have you been waiting?

I just requested the MH clusters yesterday. Haven't come yet. The Genetic Affairs ones came within a couple of hours after being requested.

MitchellSince1893
03-03-2019, 03:06 PM
I got a report but all of my father's matches disappeared from the site

MitchellSince1893
03-03-2019, 07:39 PM
I got a report but all of my father's matches disappeared from the site

Phew! My father's matches have reappeared.

mxcrowe
03-03-2019, 10:03 PM
I've been working with clusters from Genetic Affairs, which are helpful although you definitely need to analyze them.

https://blog.kittycooper.com/2018/12/automatic-clustering-from-genetic-affairs/

This seems identical but for My Heritage, and I am hoping it will add to the analysis.

Yes, glad you mentioned this. I was just going to say it seems to be similar, if not identical to Genetic Affairs clustering, which I have been planning to do (once I got past the increased security risk of giving one 3rd party company access to all my gen/gen sites...). And...it *is* the same guy: MH says they collaborated with Evert-Jan Blom, who is behind Genetic Affairs.

https://www.geneticaffairs.com/about.html

The system appears to have let me begin the clustering process for my parents. Started yesterday afternoon and no report delivered yet, but sounds like folks are waiting a day, at least. The Family Theory thing just seemed to confirm 2nd cousins I had already discovered or knew of previously, but will hopefully prove more useful in time.

It looks like everyone is trying to offer some automation to the genetic genealogy process, as this stuff from MH is similar to what Ancestry is offering with ThroughLines, which I have actually found to be pretty useful, though the UI/UE is a bit kludgy at the moment.

mxcrowe
03-04-2019, 10:53 PM
So, I got my AutoCluster results back for my mom and dad...actually quite interesting. Basically just creating triangulation groups automatically with a bit of graphical help. Also, these groups are fairly 'tight', so you're not likely to end up chasing segment holders who all share some common ancestral segment that everyone from, say, Limerick, seems to have. This was a pretty good dry run for using the full Genetic Affairs system, which will auto-cluster not just the MH testers, but also my 23andMe, Ancestry, and FTDNA folks, who may have tested at only one company, into one chart - that should be really interesting and useful, as building TGs across the various testing companies is a real PITA. Good on MH for being first with this.

Loderingo
03-05-2019, 09:09 AM
How can it create groups with Ancestry and other providers when Ancestry doesn’t have a chromosome browser?

geebee
03-05-2019, 03:17 PM
How can it create groups with Ancestry and other providers when Ancestry doesn’t have a chromosome browser?

It's because these files have all been uploaded to MyHeritage. So those results can be compared to everyone else in the MyHeritage database -- regardless of what company did the actual testing. The same thing is true for uploads to FTDNA, or to GEDmatch.

Certainly, Ancestry could have a chromosome browser as soon as they decided they were willing to provide one. They simply have no wish to provide their customers with this tool. Their reason is "privacy concerns", which while legitimate is not insurmountable; and their service will always be second-rate until they provide one.

(Although they are trying to do workarounds, in the end there really is no substitute for a chromosome browser.)

msmarjoribanks
03-05-2019, 06:17 PM
How can it create groups with Ancestry and other providers when Ancestry doesn’t have a chromosome browser?

It's not based on chromosome matches, but shared matches. The idea is that each member matches with most of the other members. The overlap in theory protects against people matching through totally different people.

msmarjoribanks
03-05-2019, 06:18 PM
How does one decipher the report? I have 10 clusters. Total 100 matches. Cluster one is 50/60 or so matches. Most are all Albanians. The latter clusters had a couple Greeks, Armenians, Turks and a Macedonians(assuming torbesh by the name).

Can you figure out how you match the closest matches within the groups?

mxcrowe
03-05-2019, 06:43 PM
How can it create groups with Ancestry and other providers when Ancestry doesn’t have a chromosome browser?

Yeah, I was thinking about this in the shower this morning...where all good thinking takes place. Short answer: I haven't actually used the full Genetic Affairs system yet, or dug into it too much, so the bulk of my exposure is a quick read and then this mini-run with just MyHeritage matches. But I think, as others have mentioned, they are able to get the 'shared matches' into it from Ancestry and cluster based on that. Obviously, that makes one wonder if those matches are truly triangulated, which I believe all the matches from the other companies will be when they auto-cluster. If Ancestry did the same thing, they clearly know the actual segments and could triangulate them, but I frankly don't know if Ancestry shared matches are triangulated matches or just matches who share a significant amount of DNA with the group. I believe the Ancestry white papers shed more light on how their algorithms actually work, but I have not yet read them all. Maybe someone here knows?

By the way, I guess this whole clustering thing is similar to what FTDNA offered with their "Matrix" tool for many years, no? Obviously, it's much nicer to have the whole thing done automatically rather than add the people to the Matrix group one by one. And, I still feel like the real value of this auto-clustering process is to do it across all your testing companies, which will create bigger clusters and include all the matches that may be spread among the various companies.

timberwolf
03-05-2019, 07:45 PM
Yeah, I was thinking about this in the shower this morning...where all good thinking takes place. Short answer: I haven't actually used the full Genetic Affairs system yet, or dug into it too much, so the bulk of my exposure is a quick read and then this mini-run with just MyHeritage matches. But I think, as others have mentioned, they are able to get the 'shared matches' into it from Ancestry and cluster based on that. Obviously, that makes one wonder if those matches are truly triangulated, which I believe all the matches from the other companies will be when they auto-cluster. If Ancestry did the same thing, they clearly know the actual segments and could triangulate them, but I frankly don't know if Ancestry shared matches are triangulated matches or just matches who share a significant amount of DNA with the group. I believe the Ancestry white papers shed more light on how their algorithms actually work, but I have not yet read them all. Maybe someone here knows?

By the way, I guess this whole clustering thing is similar to what FTDNA offered with their "Matrix" tool for many years, no? Obviously, it's much nicer to have the whole thing done automatically rather than add the people to the Matrix group one by one. And, I still feel like the real value of this auto-clustering process is to do it across all your testing companies, which will create bigger clusters and include all the matches that may be spread among the various companies.

I think I read somewhere that Gedmatch Genesis may be adding an auto clustering feature, or at least that was the rumor.

Trelvern
03-05-2019, 08:04 PM
I think I read somewhere that Gedmatch Genesis may be adding an auto clustering feature, or at least that was the rumor.

I triggered an automatic clustering process
what does it really bring?
you did it?

timberwolf
03-05-2019, 09:17 PM
I triggered an automatic clustering process
what does it really bring?
you did it?

With Myheritage?

Took about 24 hours and it clusters people into groups. Only got about 200 or the 5000 matches I had.

I would be far more interested in the more distant matches, then anything 3rd cousin or closer.

msmarjoribanks
03-06-2019, 02:26 AM
Yeah, I was thinking about this in the shower this morning...where all good thinking takes place. Short answer: I haven't actually used the full Genetic Affairs system yet, or dug into it too much, so the bulk of my exposure is a quick read and then this mini-run with just MyHeritage matches. But I think, as others have mentioned, they are able to get the 'shared matches' into it from Ancestry and cluster based on that. Obviously, that makes one wonder if those matches are truly triangulated, which I believe all the matches from the other companies will be when they auto-cluster. If Ancestry did the same thing, they clearly know the actual segments and could triangulate them, but I frankly don't know if Ancestry shared matches are triangulated matches or just matches who share a significant amount of DNA with the group. I believe the Ancestry white papers shed more light on how their algorithms actually work, but I have not yet read them all. Maybe someone here knows?

By the way, I guess this whole clustering thing is similar to what FTDNA offered with their "Matrix" tool for many years, no? Obviously, it's much nicer to have the whole thing done automatically rather than add the people to the Matrix group one by one. And, I still feel like the real value of this auto-clustering process is to do it across all your testing companies, which will create bigger clusters and include all the matches that may be spread among the various companies.

Yeah, they aren't triangulated.

Trelvern
03-06-2019, 04:24 PM
With Myheritage?

Took about 24 hours and it clusters people into groups. Only got about 200 or the 5000 matches I had.

I would be far more interested in the more distant matches, then anything 3rd cousin or closer.


I received a message telling me that a ZIP file was attached, but there was nothing at all.

ph2ter
03-06-2019, 10:50 PM
Nice graphic.
I got my clusters and the only common cluster which I have with my father is a group of people for which I found a common ancestor living in the 18th century.
All other clusters we do not have in common and for all of them I can't figure how my father or me are related to those people.

Iseid0441
03-07-2019, 09:40 PM
Mine is a one big cluster with a couple of tiny ones.

29224

FionnSneachta
03-07-2019, 10:25 PM
My family have small distinct clusters.

29225

Nqp15hhu
03-07-2019, 11:22 PM
Hi guys, I've just received my autocluster chart. I have 19 clusters, with the highest having 13 members.

How do I interpret this? I don't recognise any of the people on the Clusters.

29231

timberwolf
03-08-2019, 12:31 AM
Hi guys, I've just received my autocluster chart. I have 19 clusters, with the highest having 13 members.

How do I interpret this? I don't recognise any of the people on the Clusters.

29231

Any surnames or cities in common with your paper trail?

Nqp15hhu
03-08-2019, 12:38 AM
Any surnames or cities in common with your paper trail?

I've established a few of the clusters just by looking at the tree, but there are still a few clusters left that I cannot determine an origin to.

timberwolf
03-08-2019, 01:08 AM
I've established a few of the clusters just by looking at the tree, but there are still a few clusters left that I cannot determine an origin to.

I was able to recognize some of my clusters. But certainly not all.

Nqp15hhu
03-08-2019, 03:00 AM
I was able to recognize some of my clusters. But certainly not all.

Yeah just a few. Think one or two are Scottish, but hard to tell. How far back do these go?

timberwolf
03-08-2019, 06:12 AM
Yeah just a few. Think one or two are Scottish, but hard to tell. How far back do these go?

I did try genetic affairs and I was able to recognize the common ancestor, I think once you get to 4th cousin, it is pretty difficult.

msmarjoribanks
03-08-2019, 10:08 PM
I'm working my way through these. Some are easy to pinpoint, some are not. It's a good incentive to spend more time with my MH matches.

Osiris
03-10-2019, 09:23 AM
Mine is a one big cluster with a couple of tiny ones.

29224
Ha! My mom's has you beat. 1 cluster, 99 members. I guess this does work so well when endogamy is involved.
29267

Trelvern
03-10-2019, 05:09 PM
I was able to recognize some of my clusters. But certainly not all.


i got it
26 clusters , beautiful colours, nice graph but i don't know these people

CelticGerman
03-10-2019, 10:43 PM
A third degree cousin with mennonite background from her mother's side. 29270

CelticGerman
03-10-2019, 10:45 PM
My result: 29271

timberwolf
03-11-2019, 07:04 AM
i got it
26 clusters , beautiful colours, nice graph but i don't know these people

I take it your matches were French?

msmarjoribanks
03-11-2019, 02:41 PM
My clusters:

29284

My mom's (she has more clusters and they seem to be more balanced in size than I do):

29285

Neither of us seems to have an endogamy problem.

I've only been working with my mom's so far, as I am working on a problem on her side, and figured her matches would be the better aide. I haven't done much with MyHeritage matches in the past (they are far harder to identify on average than my Ancestry ones, although easier than 23andMe). However, many of the clusters do have people I recognize in them, so this is a helpful way to start sorting and to help figure out who the others are.

Several of my mom's clusters have only 3 members, and I am ignoring those.

geebee
03-11-2019, 06:18 PM
I have just one large cluster (55 members) and eight smaller clusters. The largest of the smaller ones has 10 members; one has 5 members; two have 4 members; and four have 3 members.

Three of the smaller clusters obviously have a close relationship to the largest cluster, and you can tell from the sort of "bands" of gray boxes extending horizontally and vertically from the smaller clusters.

29295

EDIT:

For what it's worth, while my parents and their parents don't seem to be related in any significant way, that's not true for my maternal grandfather's maternal grandparents; or for my maternal grandmother's maternal grandparents.

In both of those parts of my family, there is high endogamy. On each side, these couples were 2nd cousins. The endogamy on both sides actually continued, except that for each of my grandparents there was something that changed this pattern.

For my grandfather, it was probably the fact that he was adopted out and didn't grow up in the same place (for the most part) as his biological relatives. For my grandmother, it was the fact that she descends from her mother's 2nd husband. He was not related to my great grandmother as far as I know -- even though both spouse's fathers immigrated from the same Spanish island (Menorca). My great grandmother's 1st husband, whom she divorced, was also her 2nd cousin once removed.

What that means, among other things, is that I have several half 2nd cousins on my maternal grandmother's mother's side who are also my 5th cousins on their great grandfather's side. (He'd be something like my 2nd cousin three times removed, I think.)

I also have a 3rd cousin in this part of my family who my 3rd cousin once removed as well. He's my 3rd cousin through his father, who is my 2nd cousin once removed; and he's my 3rd cousin once removed through his mother, who is also my 3rd cousin.

I share DNA with both of his parents, and one of these segments actually triangulates -- meaning the spouses happen to share the segment with each other.

Trelvern
03-11-2019, 06:19 PM
I take it your matches were French?

My matches

1_Martial RAOULT,2_Manuel RAOULT MAESO,3_Yann Raoult-Maeso-Arnaez,4_Jacqueline Le Fur,5_Francois Hellec,6_Thomas Euzenat,7_Didier Chevreux,8_Jean-Pierre LE GOVIC,9_John Graham Fischer,10_Alexia Lind,11_Alison Adams,12_Peter Campbell,13_CLAUDE LE MANACH,14_Solen Le Lannou,15_Marie Thérèse Losfeld,16_Chloe Le Lannou-Scutt,17_Angela Khon,18_Louis Khon,19_John Hammond,20_John Dylan Luther,21_David BOURNE,22_Mckenna Aileen Clarke,23_Ruby Lea Allen,24_Camille Jansen,25_Michael E. Joseph Dufresne,26_Olwen Davies,27_Linda Matheson,28_Aileni Noyle,29_Elvira siff littau jacobsen,30_olivier bourdoulous,31_Johnny Rasmussen,32_jacques Le Merrer,33_simon Gourmelin,34_TIMOTHEE PETILLEAU,35_Sylvie Tribodet,36_james harouet,37_Edouard Ferlin,38_Josiane Coslin,39_Stéphane Michel Le Roux,40_Maryvonne Schwarz,41_John Bartelson,42_Mária Simon,43_Edmund Rybold,44_Robert Holtman,45_Abigail Sjolander,46_Arne Lennart Peterson,47_Nate Wilson,48_Bonnie Epperson,49_Lauren Brack,50_Ronan Dorvillers,51_Étienne montfort,52_jacques beleguic,53_CHRISTELE LE MENTEC,54_Lena Gallic,55_David Le solliec,56_Nicole singeot,57_Michel Carcreff,58_Juliette Marilyne SINGEOT,59_Patricia Baradeau cudennec,60_Annie Kermaidic,61_Laetitia C,62_Wanda Stead,63_adeline rançon,64_Gwendal Pondard,65_Patricia Scow,66_Maria Piotrowski,67_tim eleveld,68_Kilian Tavasier-Fauvel,69_Camille Servais,70_Raphaël Tavasier-Fauvel,71_Clea Johnson,72_Doug Johnson,73_christophe domergue,74_Sonia Geindreau,75_Moj Moj,76_Corentin Jabbour,77_Carol Sommerville,78_Stephen Whiting,79_J P,80_DAMIEN TANGUY,81_Daniel GENDRAUD,82_Carole TANGUY,83_Kathy Gorman,84_Hayden Francis Sudall,85_Анастасія Боднар


mostly French but not all
I picked a name on the list and discovered the daughter of a next-door neighbour ,now deceased, with whom I played a long time ago in my childhood!
She works in London and is a third cousin. What I have always ignored. I sent her an email, she replied.
Amazing really.

timberwolf
03-11-2019, 08:28 PM
My matches

1_Martial RAOULT,2_Manuel RAOULT MAESO,3_Yann Raoult-Maeso-Arnaez,4_Jacqueline Le Fur,5_Francois Hellec,6_Thomas Euzenat,7_Didier Chevreux,8_Jean-Pierre LE GOVIC,9_John Graham Fischer,10_Alexia Lind,11_Alison Adams,12_Peter Campbell,13_CLAUDE LE MANACH,14_Solen Le Lannou,15_Marie Thérèse Losfeld,16_Chloe Le Lannou-Scutt,17_Angela Khon,18_Louis Khon,19_John Hammond,20_John Dylan Luther,21_David BOURNE,22_Mckenna Aileen Clarke,23_Ruby Lea Allen,24_Camille Jansen,25_Michael E. Joseph Dufresne,26_Olwen Davies,27_Linda Matheson,28_Aileni Noyle,29_Elvira siff littau jacobsen,30_olivier bourdoulous,31_Johnny Rasmussen,32_jacques Le Merrer,33_simon Gourmelin,34_TIMOTHEE PETILLEAU,35_Sylvie Tribodet,36_james harouet,37_Edouard Ferlin,38_Josiane Coslin,39_Stéphane Michel Le Roux,40_Maryvonne Schwarz,41_John Bartelson,42_Mária Simon,43_Edmund Rybold,44_Robert Holtman,45_Abigail Sjolander,46_Arne Lennart Peterson,47_Nate Wilson,48_Bonnie Epperson,49_Lauren Brack,50_Ronan Dorvillers,51_Étienne montfort,52_jacques beleguic,53_CHRISTELE LE MENTEC,54_Lena Gallic,55_David Le solliec,56_Nicole singeot,57_Michel Carcreff,58_Juliette Marilyne SINGEOT,59_Patricia Baradeau cudennec,60_Annie Kermaidic,61_Laetitia C,62_Wanda Stead,63_adeline rançon,64_Gwendal Pondard,65_Patricia Scow,66_Maria Piotrowski,67_tim eleveld,68_Kilian Tavasier-Fauvel,69_Camille Servais,70_Raphaël Tavasier-Fauvel,71_Clea Johnson,72_Doug Johnson,73_christophe domergue,74_Sonia Geindreau,75_Moj Moj,76_Corentin Jabbour,77_Carol Sommerville,78_Stephen Whiting,79_J P,80_DAMIEN TANGUY,81_Daniel GENDRAUD,82_Carole TANGUY,83_Kathy Gorman,84_Hayden Francis Sudall,85_Анастасія Боднар


mostly French but not all
I picked a name on the list and discovered the daughter of a next-door neighbour ,now deceased, with whom I played a long time ago in my childhood!
She works in London and is a third cousin. What I have always ignored. I sent her an email, she replied.
Amazing really.

I take it you do get a few Anglo matches. Certainly I find with Myhertiage I get far more Continental matches then anywhere else.

I do have one Breton Women who I share a 22.4 cM segment with. I would imagine it is more to do with shared similarity between Cornwall and Brittany then actually real ancestry. But then you never quite know.

Trelvern
03-11-2019, 10:31 PM
I take it you do get a few Anglo matches. Certainly I find with Myhertiage I get far more Continental matches then anywhere else.

I do have one Breton Women who I share a 22.4 cM segment with. I would imagine it is more to do with shared similarity between Cornwall and Brittany then actually real ancestry. But then you never quite know.


I also have New Zealanders, Australians, British, Americans, Canadians ... even an Ukrainian!
And I noticed that all the French on my list are supposed to have Welsh or Irish ancestry ... I think real and fake cousins ​​are mixed here.
But ,as you said, who knows?

mildlycurly
03-18-2019, 03:47 PM
My biggest cluster has only ten members, which is extremely surprising given the majority of my matches are naturally of British or Irish origin.

Saetro
03-27-2019, 07:02 PM
I'm working my way through these. Some are easy to pinpoint, some are not. It's a good incentive to spend more time with my MH matches.
Their cunning plan has worked then.

[QUOTE=msmarjoribanks;553055]
My clusters:
29284
I haven't done much with MyHeritage matches in the past (they are far harder to identify on average than my Ancestry ones, although easier than 23andMe). However, many of the clusters do have people I recognize in them, so this is a helpful way to start sorting and to help figure out who the others are./QUOTE]

Looks similar to my DNAGedcom clustering, which was very helpful.
Apart for that big blob up the top, which for me is a heap of people with early colonial (US) origins and they have no idea where they come from in England.
I know where my people came from, but without some regionality from them, I cannot venture to suggest anything that may help them.
Other clusters have often had one person I could identify, so the others must be connected.
In one case I asked the person I knew about someone who had tested but never done anything else - certainly never checked messages - and he knew her, so at least I have a contact if I need one.

MyHeritage arranged for around 80-100 matches to be clustered, and that gave lots of small clusters, none of whom I knew.
Essentially, they were still working within that big cluster found via DNAGedcom.
The lower threshold was too high to include matches that would have provided the necessary structure.
I know that from working with my clustering at DNAGedcom - progressively reducing the lower limit until good structure appeared.
Happy that the MyHeritage approach is working for many.
Friends who have used Genetic Affairs are enjoying the extra data from that approach.

When clustering works, I find it gives more understanding of more match results more quickly than the tree/DNA combination approaches.
Simply because so many people are not making public trees and not updating them when they do.
But I need at least one of those tree/DNA combos per cluster to make progress.

I'm still waiting for the Theory of Family Relativity to reach me. (November upload.)

Basta
03-29-2019, 10:13 AM
I have clusters, but not theories. Why?

msmarjoribanks
03-29-2019, 10:30 AM
The theories are based on trees plus the cMs shared. Do you have a tree on the site? If so, maybe the matches are too far back or they don't have trees.

Phoebe Watts
03-29-2019, 05:39 PM
Having seen some good new matches, I asked an updaed auto cluster report.

I can place six out of these clusters.

The grey squares are interesting - it does look as if some of the clusters are very close together.


29562

Saetro
04-03-2019, 08:04 PM
Having seen some good new matches, I asked an updated auto cluster report.
I can place six out of these clusters.
The grey squares are interesting - it does look as if some of the clusters are very close together.

29562

Yes, your 7th and 8th clusters look related to the second.
I really enjoy superclustering (moving the clusters around) with the clustering output from DNAGedcom.
Could not see how to do it myself with the MyHeritage output.
The bright shiny tools on the html file are great, but cannot be manipulated, while the Excel file was just basic data.
Great to have predigested stuff with lots of tools, but my data does not fit the mould, so could do with a bit more massaging.

Reexamination of my output did show one maybe useful cluster, not connected to all the others.
The known person links to a relatively close fork in the tree.
So I will have to work hard in both directions.

Phoebe Watts
04-04-2019, 08:45 AM
Yes, your 7th and 8th clusters look related to the second.
I really enjoy superclustering (moving the clusters around) with the clustering output from DNAGedcom.
Could not see how to do it myself with the MyHeritage output.
The bright shiny tools on the html file are great, but cannot be manipulated, while the Excel file was just basic data.
Great to have predigested stuff with lots of tools, but my data does not fit the mould, so could do with a bit more massaging.

Reexamination of my output did show one maybe useful cluster, not connected to all the others.
The known person links to a relatively close fork in the tree.
So I will have to work hard in both directions.

I got as far as Cluster 1 looking too perfect - I can dig down to see they are all matched on one segment but I don’t see anything else in common - a bit IBS looking perhaps.

The close clusters 2,7,8 almost all have typical modern Welsh names:A maternal great-grandfather. Clusters 3 and 4 want to cluster together and I can see south Wales ancestry a couple of generations back. I have known 3rd/4th cousins in two of the small clusters and known IBS in another two.

So that’s making sense of half the clusters. I’m sure there is something to gain from some of the others too. It is a pity that there isn’t scope for more manipulation.

Donwulff
04-04-2019, 11:05 PM
Yesterday I received an e-mail that says I have "38 Theory of Family Relativity™". Going on the website shows that they've indeed added a feature that's similar to AncestryDNA's ThruLines. The thing that everybody in this thread seems to be talking about is called AutoClusters. Ahh, confusion! ;)

Nqp15hhu
06-24-2019, 11:10 PM
How are you supposed to get your head round this? I have only found two or three shared ancestors that I could identify. A lot of my matches have locked trees or a few members within their tree.

Nqp15hhu
06-24-2019, 11:48 PM
For autoclustering, all clusters for my family were generated from 25 cM to 350 cM matches.
For me, 101 matches ended up in 27 clusters. 11 matches were excluded. The largest cluster has 7 members (I don't know the connection to any of them).
My mum had 100 matches in 27 clusters with 5 excluded. The largest cluster has 9 members. I don't know the connection to any of them.
My dad had 86 matches in 26 clusters with 8 excluded. The largest cluster has 5 members. I do know the common ancestors of these matches which are my 4x great grandparents
My great aunt had 101 matches in 30 clusters with 14 excluded. The largest cluster has 7 members. I don't know the connection there either.

Some look like they make sense. However, I don't know how I'm connected to most of my matches so I'd need to spend time looking at the matches to determine what ancestors the clusters are associated with.

It's good to see that they're developing new features though that might help.

Yeah. I am struggling with this big time. They need to highlight specifically on a tree where the match is.

Some of my clusters seem to be into the 1700's?

msmarjoribanks
06-25-2019, 04:11 AM
How are you supposed to get your head round this? I have only found two or three shared ancestors that I could identify. A lot of my matches have locked trees or a few members within their tree.

It's annoying since on sites other than Ancestry I find it difficult to place enough ancestors to make the clusters as useful as can be. My Ancestry ones are super useful, but they have no chromosome information.

What I am doing is putting people in buckets, first by grandparents and then by great-grandparents and once I have enough it gets easier to figure out how new people are related. If you know some of the people I'd start by working with their clusters and identifying who you can (trace your relatives forward to help recognize names). Sometimes when you email people with "I see we are related and I think you are on X side, I will share all I have with you if you share yours with me so we can figure this out" it's more likely to get a response. I haven't emailed lots of people on MH, but the ones I have do seem interested in family history and will invite me to their tree -- more success than at 23andMe.

Nqp15hhu
06-25-2019, 11:05 AM
Several of the clusters have locked family trees and a lot of the matches are in the 1700’s.

So for example, I have a few Scots results with one Northern Irish person in the 1700’s.

An example would be someone with the surname “Wright” in Antrim.

Yeah how am I supposed to confirm that? I have absolutely no connection to Wright or Antrim. Similarly, some are in England/Scotland with no connection to NI, despite me having no recorded GB ancestors.

Unfortunately, I think most of these are too far back for me to detect. I haven’t done a family tree into the 1700’s so I can not confirm this.

firemonkey
06-25-2019, 01:21 PM
The thing that gets me is that most of these clusters are not a solid block .