Page 2 of 3 FirstFirst 123 LastLast
Results 11 to 20 of 21

Thread: How accurate is GEDmatch really?

  1. #11
    Registered Users
    Posts
    1,396
    Sex
    Location
    America
    Ethnicity
    North & Ionian Seas
    Nationality
    American

    England Italy Germany Scotland
    Quote Originally Posted by digital_noise View Post
    can you provide a source this? For as far as I have been using GedMatch and Genesis the results are identical
    the only source I have is my own data which changed a bit in May 2018, but has since gone back. the change was at the start of the Genesis migration. They still have a statement about launching a new algorithm but they do not say when.

    the Admixture Proportions by Chromosome is still messed up/incomplete so I'm not sure they've gotten anything working right other than consolidation kits onto one page i.e. Migration.
    Last edited by JerryS.; 02-11-2019 at 10:21 PM.

  2. #12
    Registered Users
    Posts
    743
    Sex
    Location
    Bay Area
    Nationality
    American
    Y-DNA
    E-BY14160
    mtDNA
    J1c2p

    Italy Two Sicilies England Sweden Netherlands
    Quote Originally Posted by JerryS. View Post
    the only source I have is my own data which changed a bit in May 2018, but has since gone back. the change was at the start of the Genesis migration. They still have a statement about launching a new algorithm but they do not say when.

    the Admixture Proportions by Chromosome is still messed up/incomplete so I'm not sure they've gotten anything working right other than consolidation kits onto one page i.e. Migration.
    where is this statement? I cannot find anything on their page mentioning this.
    nMonte K36
    Fit = 1.87
    NE England 47.5%
    IT Calabria 33.3%
    NL_Friesland 13.3%
    SV_Svealand 5.83%

    G25
    Fit=2.03
    Eng 61.67%
    IT South 30%
    Sweden 8.33%


    E-V13 => E-PH1246 => E-BY14160
    Antonio Reale born circa 1710, Ciminą Italy

  3. #13
    Registered Users
    Posts
    1,396
    Sex
    Location
    America
    Ethnicity
    North & Ionian Seas
    Nationality
    American

    England Italy Germany Scotland
    Quote Originally Posted by digital_noise View Post
    where is this statement? I cannot find anything on their page mentioning this.
    **** copied and pasted just now.

    The Genesis Algorithm

    For several years, GEDmatch has provided genetic genealogists, both beginners and experts, the ability to search for matches among kits in their database without regard to vendor. Also, GEDmatch has provided a rich suite of analysis programs allowing users to dig deeply into the genetic details of their matches, enhance the reports from their vendors, and even pursue their own original research ideas. Our algorithms are evolving to extract the most trustworthy and meaningful matching information possible using the markers common to pairs of kits even though sometimes limited.

    Unfortunately, all too often, kits appear to share a DNA segment purely by chance. To combat this confusing phenomenon, we recently have developed a reliability measure that allows users to assess the quality of a matching segment in an intuitively appealing fashion. We also use the measure to guide our matching algorithms as they wring the greatest amount of useful information possible from the markers common to pairs of kits.

    If we could assume that marker characteristics were uniform in all regions within chromosomes, we could use a "one size fits all" requirement for matching segments as is sometimes done. Unfortunately, the relevant characteristics vary widely. Some long segments with few markers may be accidental matches. Some marker rich short segments are often discarded although they are profoundly non-random.

    Using the characteristics of each and every marker in a segment, we compute the expected number of purely chance matches to it to be found in the database. That number is then used to classify the segment into one of several levels reflecting the likelihood that the random matches may overwhelm the real ones. When a user executes a one-to-many search or a one-to-one comparison specifying a minimum segment length, the display can then include an estimate of validity for each segment found.

    One can assume those segments designated to be valid are the result of a DNA inheritance process rather than mere chance. Questions may still remain about how far back shared DNA originates, but a confounding factor has been removed.

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

     digital_noise (02-11-2019)

  5. #14
    Registered Users
    Posts
    743
    Sex
    Location
    Bay Area
    Nationality
    American
    Y-DNA
    E-BY14160
    mtDNA
    J1c2p

    Italy Two Sicilies England Sweden Netherlands
    I see. I could be mistaken but thats for matches, not the calculators?
    nMonte K36
    Fit = 1.87
    NE England 47.5%
    IT Calabria 33.3%
    NL_Friesland 13.3%
    SV_Svealand 5.83%

    G25
    Fit=2.03
    Eng 61.67%
    IT South 30%
    Sweden 8.33%


    E-V13 => E-PH1246 => E-BY14160
    Antonio Reale born circa 1710, Ciminą Italy

  6. #15
    Registered Users
    Posts
    1,396
    Sex
    Location
    America
    Ethnicity
    North & Ionian Seas
    Nationality
    American

    England Italy Germany Scotland
    Quote Originally Posted by digital_noise View Post
    I see. I could be mistaken but thats for matches, not the calculators?
    I don't know. I figured if your matches change maybe other things do too to some small extent... I guess to answer the OP's question with my opinion, it is close but no cigar.

  7. #16
    The thing I like about Gedmatch is the oracles and ancestral populations they give you by genetic distance. Also, I remember user LukaszM was kind enough to add me to K13 PCA which he made. I plotted between South Dutch and English, which was very cool information. I do think Gedmatch is has problems though, mainly some admixture they give people which can be either flat out inaccurate or from ancient times.

  8. The Following User Says Thank You to ianz91 For This Useful Post:

     JerryS. (02-12-2019)

  9. #17
    Registered Users
    Posts
    189
    Sex
    Location
    Canada
    Ethnicity
    European
    Nationality
    Typical Canadian Mutt
    Y-DNA
    R-DF27 -> FGC34856
    mtDNA
    J1c2e

    Ireland Ukraine England France Canada
    Matching cMs are more accurate than AncestryDNA, which uses TIMBER and strips out valid matching segments labelling them as "too matchy". GEDMATCH is still the only game in town where you can phase your own data, and besides 23andme, the only one that has chromosome painting capabilities. But that being said, the Ethnicity calculators are very old and outdated; I'm sure they're trustworthy at the continental level, but anything more than that they cannot compete with today's massive reference panels and new arrays of tested SNPs.
    Mum = 50% Irish, 25% Ukrainian, 25% Romanian
    Dad = 40% French-Canadian, 10% Irish, 50% English
    Big Y + YTree.net = R M269 -> DF27 -> Z195 -> FGC34881 -> FGC34865 (SW France; ~500 ybp)
    FTDNA mtDNA Full Sequence = J1c2e
    Most Distant Known Ancestor = Jean Moreau b. 1630s Parthenay, Deux-Sčvres, France
    Surnames = Welch, Chibry, Moreau, Todd, Anderson, Bedford, Joncas, Basaraba

  10. The Following User Says Thank You to fostert For This Useful Post:

     Judith (02-18-2019)

  11. #18
    Registered Users
    Posts
    516
    Sex
    Ethnicity
    Kurd from Turkey (Dersim)
    Y-DNA
    E-L29
    mtDNA
    H

    Kurdistan Turkey Kurdistan Kingdom Germany Bundeswappen
    gedmatch is good
    Kurdish people gallery/thread - Hidden Content
    Hidden Content

    another link with more photos/videos !! Hidden Content

    our anthroforum : Hidden Content join us Hidden Content


  12. #19
    Gold Member Class
    Posts
    884
    Sex
    Location
    Brisbane
    Nationality
    Australian
    Y-DNA
    T-P322 (T1a2b1)
    mtDNA
    H6a1

    Australia Cornwall England Scotland Germany Poland
    Quote Originally Posted by digital_noise View Post
    I see. I could be mistaken but thats for matches, not the calculators?
    Last time I looked at the opening panel with the transfer progress report (a couple of days ago), it said something about the ethnicity calculators yet to be completed.
    And aren't they constructed by independent developers, not the main site people?
    So who knows when they will be up again.
    And with the tiny amount of overlap from 23andMe v5, that chip might be really difficult to accommodate.
    The others should be OK.

  13. #20
    Registered Users
    Posts
    21

    Gedmatch seems outdated and flawed to me. Very inconsistent calculators and oracles with some questionable reference samples. If they smooth these errors then I think it will be more effective. I think the nMonte tool is actually better than Gedmatch.

Page 2 of 3 FirstFirst 123 LastLast

Similar Threads

  1. How accurate is your gedmatch eye color prediction?
    By Kart in forum Anatomy and Physiology
    Replies: 35
    Last Post: 12-26-2018, 09:29 PM
  2. Best or most accurate Gedmatch Admixture?
    By Jutlander in forum Autosomal (auDNA)
    Replies: 53
    Last Post: 12-02-2018, 02:20 PM
  3. Replies: 1
    Last Post: 09-04-2018, 09:55 PM
  4. Replies: 41
    Last Post: 05-14-2018, 03:24 AM
  5. How accurate are the 'One-to-many' matches on GEDMatch
    By Mehrdad in forum Autosomal (auDNA)
    Replies: 1
    Last Post: 03-04-2014, 11:17 AM

Posting Permissions

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