Announcement

Collapse
No announcement yet.

Different Companies, Different Results

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Different Companies, Different Results

    I get that different companies have different reference populations and algorithms, but this is a little ridiculous (and also very confusing).
    I tested with three different companies and used third-party analyses, however I get drastically different results each time. Here's an example of the range of my data:
    Central Europe: 0-73%
    Great Britain/Ireland: 11-70%

    Even the non-European ones are inconsistent. I have some Middle Eastern over here, some Native American over there, etc.

    Comparing my raw 23andMe data and Ancestry data in GEDMatch chromosome compare shows that they don't match, but it's all my DNA.
    I also ran this data through Dna.Land to see what it came up with, but I still got drastically different results.
    What gives? Anyone else get extremely different results between companies?

    With the information I'm getting, I've decided that I should just pick and choose what I like the best

  • #2
    Originally posted by Razska View Post
    Comparing my raw 23andMe data and Ancestry data in GEDMatch chromosome compare shows that they don't match, but it's all my DNA.
    Wait, so you did a one to one comparison with both your kits and they don't match at all?

    Comment


    • #3
      Originally posted by Razska View Post
      I get that different companies have different reference populations and algorithms, but this is a little ridiculous (and also very confusing).
      I tested with three different companies and used third-party analyses, however I get drastically different results each time. Here's an example of the range of my data:
      Central Europe: 0-73%
      Great Britain/Ireland: 11-70%
      That's still European and western Europeans and some Central Europeans share a lot of ancient ancestors which is why there isn't a consistency there.

      Originally posted by Razska View Post
      Even the non-European ones are inconsistent. I have some Middle Eastern over here, some Native American over there, etc.
      If the Native American is 2.0% or less at Gedmatch but non-existent at AncestryDNA, 23andme, and FTDNA then it is just noise. If it is less than 3% at Gedmatch it could still be noise but you would have to get some relatives tested and see which segments are identified as NA and see how those segments show up in the various Gedmatch calculators.

      Originally posted by Razska View Post
      Comparing my raw 23andMe data and Ancestry data in GEDMatch chromosome compare shows that they don't match, but it's all my DNA.
      I also ran this data through Dna.Land to see what it came up with, but I still got drastically different results.
      What gives? Anyone else get extremely different results between companies?

      With the information I'm getting, I've decided that I should just pick and choose what I like the best
      DNA.Land have fewer samples in their reference dataset than the testing companies and the company is still new. They have a long way to go to get the results refined.

      Comment


      • #4
        Sometimes you also have to do a little mind-reading to get at exactly what they mean by their varying terminology.

        For example, when I ran my kit through DNA.Land, it said that I had a significant ITALIAN (Tuscan/Bergamo) admixture. That was completely unexpected. Way out of left field. I have zero Italian ancestry whatsoever, and no ancient Mediterranean Neolithic admixture that could amount to that quantity.

        But...

        When you look on a map, Bergamo and Tuscany are in far northern Italy. I have a fair number of ancestors from Switzerland and the coastal southeastern region of France. So undoubtedly that's what that "Italian" admixture represents.

        Likewise, one of the Eurogenes tools showed me with a lot of "Baltic" admixture. It really was just a poor way of mixing/describing Scandinavian elements of my autosome that other tools teased out more carefully.

        So try not to look at actual label descriptors. Instead, look at broad "pieces of the pie", whatever they're labelled. Then try to figure out how they might match up.

        Comment


        • #5
          You shouldn't have to do that. DNA.Land needs to improve a lot.

          Comment


          • #6
            Originally posted by Armando View Post
            You shouldn't have to do that. DNA.Land needs to improve a lot.
            Absolutely agreed.

            It's still a very newborn project though, and it's free, so I didn't really care. I have better info from the original testing companies anyway.

            Comment

            Working...
            X