Author Topic: Ancestry ethnicity results updated - accuracy improved  (Read 1576 times)

Offline sallyyorks

  • RootsChat Aristocrat
  • ******
  • Posts: 1,122
    • View Profile
Re: Ancestry ethnicity results updated - accuracy improved
« Reply #9 on: Thursday 13 September 18 11:49 BST (UK) »


So, still open to further improvement, but the estimates are getting better. Presumably as the number of British testers increase, the estimates will improve further...

It wouldn't make any difference how many people take the test or where they are from.
For the purpose of 'ethnicity', your test is not compared with other testers.

RootsChat is the busiest, largest free family history forum site in the country. It is completely free to use. Register now.
Also register instantly with Facebook or Twitter (and other social networks). Start your genealogy search now.


Offline Vance Mead

  • RootsChat Member
  • ***
  • Posts: 162
    • View Profile
Re: Ancestry ethnicity results updated - accuracy improved
« Reply #10 on: Thursday 13 September 18 12:05 BST (UK) »
Quote

It wouldn't make any difference how many people take the test or where they are from.
For the purpose of 'ethnicity', your test is not compared with other testers.


Who are they comparing with?
Mead - Herts, Bucks, Essex
Pontifex - Bucks
Goldhurst - London, Middx, Herts
Kellogg/Kelhog - Essex, Cambs

RootsChat is the busiest, largest free family history forum site in the country. It is completely free to use. Register now.
Also register instantly with Facebook or Twitter (and other social networks). Start your genealogy search now.


Offline sugarfizzle

  • RootsChat Veteran
  • *****
  • Posts: 971
    • View Profile
Re: Ancestry ethnicity results updated - accuracy improved
« Reply #11 on: Thursday 13 September 18 12:10 BST (UK) »


So, still open to further improvement, but the estimates are getting better. Presumably as the number of British testers increase, the estimates will improve further...

It wouldn't make any difference how many people take the test or where they are from.
For the purpose of 'ethnicity', your test is not compared with other testers.

I think I might disagree with that, sallyorks, though I don't know for sure.

https://support.ancestry.com/s/article/Viewing-Ethnicity-Results-from-AncestryDNA-US-1460088591488-2556?ui-force-components-controllers-recordGlobalValueProvider.RecordGvp.getRecord=1&r=4

Reference panel

The AncestryDNA® reference panel is a database of 3,000 DNA samples from people selected for their deep regional roots and documented family trees. To determine your ethnic breakdown, we survey your DNA at over 700,000 locations and determine how much ethnicity you share with the people from our panel in each region.

Regards Margaret
STEER, mainly Surrey, Kent; PINNOCKS/HAINES, Gosport, Hants; BARKER, mainly Broadwater, Sussex; Gosport, Hampshire; LAVERSUCH, Micheldever, Hampshire; WESTALL, London, Reading, Berks; HYDE, Croydon, Surrey; BRIGDEN, Hadlow, Kent and London; TUTHILL/STEPHENS, London
WILKINSON, Leeds, Yorkshire and Liverpool; WILLIAMSON, Liverpool; BEARE, Yeovil, Somerset; ALLEN, Kent and London; GORST, Liverpool; HOYLE, mainly Leeds, Yorkshire

Census Information is Crown Copyright, from www.nationalarchives.go

Offline sallyyorks

  • RootsChat Aristocrat
  • ******
  • Posts: 1,122
    • View Profile
Re: Ancestry ethnicity results updated - accuracy improved
« Reply #12 on: Thursday 13 September 18 12:19 BST (UK) »
For the purpose of 'ethnicity estimates', they compare your test  with their 'reference' samples. These are the very small number of stock samples they hold, and they are not samples from other customers.

Offline Redman45

  • RootsChat Extra
  • **
  • Posts: 3
  • Census information Crown Copyright, from www.nationalarchives.gov.uk
    • View Profile
Re: Ancestry ethnicity results updated - accuracy improved
« Reply #13 on: Thursday 13 September 18 12:36 BST (UK) »
Hi Margaret,

Yes i suppose they are more accurate in a way I'm mainly English, Welsh and Irish.

Offline sugarfizzle

  • RootsChat Veteran
  • *****
  • Posts: 971
    • View Profile
Re: Ancestry ethnicity results updated - accuracy improved
« Reply #14 on: Thursday 13 September 18 12:42 BST (UK) »
For the purpose of 'ethnicity estimates', they compare your test  with their 'reference' samples. These are the very small stock samples they hold, and they are not samples from other customers.

Agreed, but the 3000 people they compare you to are other customers. As their database increases I think it possible, though again I don't know for sure, that their 3,000 people might change, or that they will look at them again more closely. 

They have obviously re-examined them recently to change the estimates so dramatically.

Either way, I am not particularly interested in the ethnicity estimates, as I have said, and never will be.

It is the matches I am interested in. As a lot of people do it for ethnicity, if that improves further and if it becomes more reliable a lot more people might get tested, over 10 million with Ancestry at present.

That means more matches for me, and more matches for you, and maybe increased accuracy of ethnicity for those that are interested.

Regards Margaret
STEER, mainly Surrey, Kent; PINNOCKS/HAINES, Gosport, Hants; BARKER, mainly Broadwater, Sussex; Gosport, Hampshire; LAVERSUCH, Micheldever, Hampshire; WESTALL, London, Reading, Berks; HYDE, Croydon, Surrey; BRIGDEN, Hadlow, Kent and London; TUTHILL/STEPHENS, London
WILKINSON, Leeds, Yorkshire and Liverpool; WILLIAMSON, Liverpool; BEARE, Yeovil, Somerset; ALLEN, Kent and London; GORST, Liverpool; HOYLE, mainly Leeds, Yorkshire

Census Information is Crown Copyright, from www.nationalarchives.go

Offline JaneyH_104

  • RootsChat Extra
  • **
  • Posts: 94
  • My great-grandmother, Elizabeth Dyson
    • View Profile
Re: Ancestry ethnicity results updated - accuracy improved
« Reply #15 on: Thursday 13 September 18 13:08 BST (UK) »
There are two main changes Ancestry have made. First, the reference panel that they compare your DNA has been increased from 3,000 to 16,000. Second, the way in which they do the calchas changed.  Previously, they made comparisons at individual locations, but now they examine longer strands of DNA.

There’s a really good blog article by Donna Rutherford here: http://donnarutherford.com/finally-the-ancestry-dna-update-weve-all-been-waiting-for/ which includes links to Ancestry’s omen article and longer technical paper.
BOWDLER - Forest of Dean & Devon, DYSON, ENTWISTLE & TOWNEND - Huddersfield, CLARKE - Dorset, SCOBLE - Devon, HOUGH, COPE & WHITTAKER - Cheshire, BRACHER - Wiltshire, DENNISS - Herts/Hunts, SQUIRE - Hunts/Beds, BROWN - Herts/Beds

Offline sallyyorks

  • RootsChat Aristocrat
  • ******
  • Posts: 1,122
    • View Profile
Re: Ancestry ethnicity results updated - accuracy improved
« Reply #16 on: Thursday 13 September 18 13:11 BST (UK) »
For the purpose of 'ethnicity estimates', they compare your test  with their 'reference' samples. These are the very small stock samples they hold, and they are not samples from other customers.

Agreed, but the 3000 people they compare you to are other customers. As their database increases I think it possible, though again I don't know for sure, that their 3,000 people might change, or that they will look at them again more closely. 

They have obviously re-examined them recently to change the estimates so dramatically.

Either way, I am not particularly interested in the ethnicity estimates, as I have said, and never will be.

It is the matches I am interested in. As a lot of people do it for ethnicity, if that improves further and if it becomes more reliable a lot more people might get tested, over 10 million with Ancestry at present.

That means more matches for me, and more matches for you, and maybe increased accuracy of ethnicity for those that are interested.

Regards Margaret


They have 43 'regions' and the world wide reference samples of 3,000 divided by 43 gives an average of 69 samples per region. It wasn't that long ago that these reference samples were as low as about 25 for a 'region' such as England or Scotland . That means your sample was probably only being compared to about 25 samples for each 'region'.
These small numbers of reference samples are from people who ancestry claim to know the 'ethnicity' of. But the problems with that as I see it are ...

Most people, no matter how good they are at genealogy, cannot trace most of their ancestry back further than about the mid to late 17th century. Not comparing myself to professional genealogists, but in my own tree, I cannot trace ancestors on most lines back before the start of industrialisation. A period that saw a great deal of migration and inter marriage within the British Isles. Some I cannot even find before the 19th century. The records either never existed in the first places, have not survived over time or are not available.

Even if you could, by magic, get back 2,000 years, even that far back, people still moved around and intermarried. They were displaced by wars, famines and so on and so they migrated to another area, often nearby yes but different to dna 'regions'. The Romans conquered half of Europe and moved armies around for example. During the Norman Conquest, especially 1069 in the north of England, more displacement and so on

We cannot really ever know how accurate these 'reference' samples are and we are not party to these samples trees. Someone whose test is used in these reference samples could have proven, by the company, ancestors in parish records in the 17th century in a certain 'region' but that not does mean they were originally from that specific place, if you could trace them in paper records further back they could have come from somewhere in another 'region'.

The people of north western Europe are basically the same ethnicity. North France, Belgium, England, Ireland Scotland and Wales are the same people. They have been mixing and intermarrying for thousands of years. Yet ancestry claim they can split those peoples up into ethnic 'regions'. To be fair they are probably under a great deal of pressure from their customers who demand it, especially  in the USA

Yes the 'matches' are worth pursuing, assuming the tree you match with is accurate, but the ethnicity part of the test isn't really to be taken seriously

Offline sallyyorks

  • RootsChat Aristocrat
  • ******
  • Posts: 1,122
    • View Profile
Re: Ancestry ethnicity results updated - accuracy improved
« Reply #17 on: Thursday 13 September 18 13:32 BST (UK) »
There are two main changes Ancestry have made. First, the reference panel that they compare your DNA has been increased from 3,000 to 16,000. Second, the way in which they do the calchas changed.  Previously, they made comparisons at individual locations, but now they examine longer strands of DNA.

There’s a really good blog article by Donna Rutherford here: http://donnarutherford.com/finally-the-ancestry-dna-update-weve-all-been-waiting-for/ which includes links to Ancestry’s omen article and longer technical paper.


Even if the reference samples are now higher in number, and I suppose this must be an improvement of sorts, but this is still a USA companies idea of what makes someones ethnicity 'Irish' or 'English' etc. They are subjective opinions and also I am sure the companies are very much influenced by their American customers ideas about European history and by their demands to split North Western Europeans into boxes of 'regions', but these companies regions are loosely based on modern political maps
You see blogs where testers talk about having 'Celt', 'Viking' or 'Anglo-Saxon/Germanic' ancestors when these terms bear no real relation to ethnicity at all.