Page 1 of 4 123 ... LastLast
Results 1 to 10 of 34

Thread: Far out unscaled populations - anyone else get them?

  1. #1
    Registered Users
    Posts
    559
    Sex
    Ethnicity
    English and Irish
    Nationality
    British

    England Ireland Connacht East Anglia England North of England United Kingdom Ireland

    Far out unscaled populations - anyone else get them?

    My scaled results are pretty standard. I am British/Irish and not Norwegian, but Norway isn’t far away. Small percentages of other populations pop up, but could be dismissed as noise.

    Distance: 1.8162% / 0.01816228
    40.8 Irish
    28.0 Norwegian
    20.2 Swedish
    8.0 Basque_French
    2.2 Cossack_Kuban
    0.4 Papuan
    0.2 Bolivian_Cochabamba
    0.2 Sakha

    (Iceland removed)

    But when I use unscaled something odd happens:

    Distance: 1.1454% / 0.01145402
    53.8 Irish
    28.8 Norwegian
    5.2 Spanish_Asturias
    3.8 Cossack_Kuban
    3.4 Japanese
    2.6 French_Pas-de-Calais
    1.6 Sakha
    0.4 Mari
    0.4 Papuan
    (Iceland removed)

    Note: I may have some as yet unconfirmed Southern European ancestry too (which was on my 23andme).

    I managed to get the Japanese to lower a bit by removing Britain, France and Spain, but it didn’t go away.

    Distance: 1.2663% / 0.01266297 | ADC: 0.25x RC
    47.4 Norwegian
    45.6 Dutch
    3.2 Japanese
    2.4 Berber_Tunisia_Chen
    1.4 Sakha
    (Western Europe removed)

    Now I don’t think I have any East Asian ancestry. It’s possible that I could have traces from colonial times, but I feel like 23andme or Wegene would have picked this up. On Wegene I do get 0.01% Chinese and 0.29% Korean, but there’s 0 Asian at all on 23andme.

    Anyone else get some far out populations? What’s your view on using unscaled? Could this be very distant colonial ancestry, or is it some background ‘Asian’ found in Europe due to old migration.
    Last edited by lana6765; 11-12-2020 at 01:21 PM.

  2. #2
    Global Moderator
    Posts
    3,760
    Sex
    Location
    Vissaiom
    Ethnicity
    Portuguese highlander
    Y-DNA (P)
    E-Y31991>FT17866
    mtDNA (M)
    H20 (xH20a)

    Asturias Galicia Portugal 1143 Portugal 1485 Portugal Order of Christ PortugalRoyalFlag1830
    It's just overfitting, this has been discussed to death in other topics
    YDNA E-Y31991>PF4428>Y134097>Y134104>Y168273>FT17866 (TMRCA ~1100AD) - Domingos Rodrigues, b. circa 1690 Hidden Content , Viana do Castelo, Portugal - Stonemason, miller.
    mtDNA H20 - Monica Vieira, b. circa 1700 Hidden Content , Porto, Portugal

    Hidden Content
    Global25 PCA West Eurasia dataset Hidden Content

    [1] "distance%=1.6007"

    Ruderico

    NW_Iberia_IA,80.4
    Berber_EMA,11
    Roman_Colonial,8.6

  3. The Following 2 Users Say Thank You to Ruderico For This Useful Post:

     BalkanKiwi (11-12-2020),  JMcB (11-12-2020)

  4. #3
    Suspended Account
    Posts
    4,850
    Sex

    A guide to proper ancestry modeling with the G25 is available here.

    https://eurogenes.blogspot.com/2018/...-davidski.html

  5. The Following 2 Users Say Thank You to Generalissimo For This Useful Post:

     jadegreg (11-12-2020),  JMcB (11-12-2020)

  6. #4
    Registered Users
    Posts
    559
    Sex
    Ethnicity
    English and Irish
    Nationality
    British

    England Ireland Connacht East Anglia England North of England United Kingdom Ireland
    Quote Originally Posted by Ruderico View Post
    It's just overfitting, this has been discussed to death in other topics
    It has?

    How do I get it to stop over fitting?

  7. #5
    Registered Users
    Posts
    2,777
    Sex
    Location
    America
    Ethnicity
    North & Ionian Seas
    Nationality
    American
    Y-DNA (P)
    I1 (P109)

    England Italy Germany Scotland
    Quote Originally Posted by lana6765 View Post
    It has?

    How do I get it to stop over fitting?
    You can't because its based on the population sources used in any particular model. Like I usually get half of a percent of Australian Papua new Guinea.

  8. #6
    Global Moderator
    Posts
    3,760
    Sex
    Location
    Vissaiom
    Ethnicity
    Portuguese highlander
    Y-DNA (P)
    E-Y31991>FT17866
    mtDNA (M)
    H20 (xH20a)

    Asturias Galicia Portugal 1143 Portugal 1485 Portugal Order of Christ PortugalRoyalFlag1830
    Quote Originally Posted by lana6765 View Post
    It has?

    How do I get it to stop over fitting?
    Yes, in multiple topics over the years. It's hard to avoid G25+nMonte from overfitting because some modern populations are just too similar from one another, notice you still get Norwegian and Swedish regardless of the type of data you use. You can remove references that make no sense for your ethnic background (ie Japanese, Papuan) from the input, and keep the models relatively simple, that's what I do
    YDNA E-Y31991>PF4428>Y134097>Y134104>Y168273>FT17866 (TMRCA ~1100AD) - Domingos Rodrigues, b. circa 1690 Hidden Content , Viana do Castelo, Portugal - Stonemason, miller.
    mtDNA H20 - Monica Vieira, b. circa 1700 Hidden Content , Porto, Portugal

    Hidden Content
    Global25 PCA West Eurasia dataset Hidden Content

    [1] "distance%=1.6007"

    Ruderico

    NW_Iberia_IA,80.4
    Berber_EMA,11
    Roman_Colonial,8.6

  9. The Following 4 Users Say Thank You to Ruderico For This Useful Post:

     jadegreg (11-12-2020),  JerryS. (11-12-2020),  JMcB (11-12-2020),  jstephan (11-12-2020)

  10. #7
    Registered Users
    Posts
    1,101
    Sex
    Y-DNA (P)
    R1a1a1b1a2b3
    mtDNA (M)
    U5a2b

    Polish–LithuanianCommonwealth Bulgaria Byzantine Empire Russian Federation Israel Russia Imperial
    Quote Originally Posted by lana6765 View Post
    It has?

    How do I get it to stop over fitting?
    I remove populations. For example on ALL models I have Baltic, then my Southern ancestry increases to 30-40% which is really 13%, which accurately shows after deleting these populations

  11. #8
    Registered Users
    Posts
    2,777
    Sex
    Location
    America
    Ethnicity
    North & Ionian Seas
    Nationality
    American
    Y-DNA (P)
    I1 (P109)

    England Italy Germany Scotland
    I too suffer from these over fitting issues often giving me far reaches of various populations. I think in the future (like Ruderico) I'm just going to scan through the source populations and remove everything that's not around the North Sea and Mediterranean Sea and see how that works out for me. No need to have Russian or Polish or Australian Papua sample....
    Last edited by JerryS.; 11-12-2020 at 05:05 PM. Reason: Spelling

  12. The Following User Says Thank You to JerryS. For This Useful Post:

     JMcB (11-12-2020)

  13. #9
    Registered Users
    Posts
    559
    Sex
    Ethnicity
    English and Irish
    Nationality
    British

    England Ireland Connacht East Anglia England North of England United Kingdom Ireland
    Quote Originally Posted by CyrylBojarski View Post
    I remove populations. For example on ALL models I have Baltic, then my Southern ancestry increases to 30-40% which is really 13%, which accurately shows after deleting these populations
    But if I remove all Asian and all Scandinavian, things just get too Ireland and then too Russia.

    Target: lana6765
    Distance: 1.1873% / 0.01187324
    75.2 Irish
    12.4 French_Pas-de-Calais
    6.0 Cossack_Kuban
    2.4 Sakha
    2.2 Dutch
    0.4 Berber_MAR_TIZ
    0.4 Mari
    0.4 Nivkh
    0.4 Papuan
    0.2 Spanish_Asturias

    I guess if I remove Russia things get less Irish. But then I still don't get English.

    Target: lana6765
    Distance: 1.2319% / 0.01231889
    64.4 Irish
    18.8 French_Pas-de-Calais
    10.4 Dutch
    4.4 Estonian
    0.8 Berber_Tunisia_Chen
    0.6 Koinanbe
    0.4 Bantu_Kenya
    0.2 Bolivian_Cochabamba

    And if I only include the places I know I have ancestry from, or think I might, things get even more Irish than ever, with very little English at all!

    Without French:
    Target: lana6765
    Distance: 1.3533% / 0.01353300
    83.4 Irish
    9.6 English
    4.8 Spanish_Cataluna
    2.2 Spanish_Eivissa

    With French:
    Target: lana6765
    Distance: 1.2797% / 0.01279720
    72.0 Irish
    27.0 French_Pas-de-Calais
    1.0 French_Seine-Maritime

    Also, isn't this sort of confirmation bias?
    Last edited by lana6765; 11-12-2020 at 05:54 PM.

  14. The Following User Says Thank You to lana6765 For This Useful Post:

     Nqp15hhu (11-12-2020)

  15. #10
    Registered Users
    Posts
    2,777
    Sex
    Location
    America
    Ethnicity
    North & Ionian Seas
    Nationality
    American
    Y-DNA (P)
    I1 (P109)

    England Italy Germany Scotland
    Quote Originally Posted by lana6765 View Post
    But if I remove all Asian and all Scandinavian, things just get too Ireland and then too Russia.

    Target: lana6765
    Distance: 1.1873% / 0.01187324
    75.2 Irish
    12.4 French_Pas-de-Calais
    6.0 Cossack_Kuban
    2.4 Sakha
    2.2 Dutch
    0.4 Berber_MAR_TIZ
    0.4 Mari
    0.4 Nivkh
    0.4 Papuan
    0.2 Spanish_Asturias

    I guess if I remove Russia things get less Irish. But then I still don't get English.

    Target: lana6765
    Distance: 1.2319% / 0.01231889
    64.4 Irish
    18.8 French_Pas-de-Calais
    10.4 Dutch
    4.4 Estonian
    0.8 Berber_Tunisia_Chen
    0.6 Koinanbe
    0.4 Bantu_Kenya
    0.2 Bolivian_Cochabamba

    And if I only include the places I know I have ancestry from, or think I might, things get even more Irish than ever, with very little English at all!

    Target: lana6765
    Distance: 1.3533% / 0.01353300
    83.4 Irish
    9.6 English
    4.8 Spanish_Cataluna
    2.2 Spanish_Eivissa

    Also, isn't this sort of confirmation bias?
    I don't know your ethnic background but you have to remember there is a lot of overlap in regional populations. Irish Scottish English Welsh Can be blended into one depending on the mode and the samples used. The same for Polish and Baltic and sometimes even Russian and other Eastern European countries. North German and scandinavian can be blended. Iberian Italian and Greek can be blended or divided. It might be more important to look for a North South or East West split rather then particular populations.

  16. The Following User Says Thank You to JerryS. For This Useful Post:

     jadegreg (11-12-2020)

Page 1 of 4 123 ... LastLast

Similar Threads

  1. MENA K18 - Unscaled
    By RickW85 in forum Autosomal (auDNA)
    Replies: 55
    Last Post: 01-21-2021, 03:59 PM
  2. A little calculator (one more!) for Europeans with G25 UNSCALED
    By anglesqueville in forum Autosomal (auDNA)
    Replies: 142
    Last Post: 11-13-2020, 09:08 PM
  3. Germanics in G25 (unscaled)
    By Finn in forum Autosomal (auDNA)
    Replies: 123
    Last Post: 09-06-2020, 07:28 PM
  4. 14 Populations Worldwide - Unscaled Calculator
    By RickW85 in forum Autosomal (auDNA)
    Replies: 78
    Last Post: 09-02-2020, 08:31 PM
  5. Unscaled vs scaled
    By firemonkey in forum Autosomal (auDNA)
    Replies: 0
    Last Post: 01-11-2020, 04:26 PM

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •