Announcement

Collapse
No announcement yet.

myOrigins 2.0 algorithm

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

  • myOrigins 2.0 algorithm


    Hi everyone,

    I have bought a test for me, my father and my mother.

    These were the results:
    https://www.dropbox.com/s/gozh19y8po...%20familia.png

    After months of discussion in foruns like anthrogenica.com, I have become more certain that there's something wrong with at least one of the results. Presumably, the results of my father. I have used the raw DNA data provided by you to run another calculators and there's no such a discrepancy between my result and the result of my father.

    Examples:
    https://www.dropbox.com/s/8moph88yq2aupl6/K13.png
    https://www.dropbox.com/s/ap3fmjzflbd6g3f/K15.png

    We are all portuguese (without non-portuguese known ancestry) and myOrigins says that my father has 44% West and Central Europe. This by itself is very very strange. More strange it becomes when we see that I have 0% West and Central Europe.
    Another thing is that I have 17% East Europe and my parents have 0 or so.

    Some people in the foruns told me to ask you for a refund. Is this a problem with the myOrigins 2.0 algorithm? What do you think?

  • #2
    First item, this is a public forum, not FTDNA customer service. I don't work for FTDNA.

    Second item, I'm going to invoke AncestryDNA for an example:

    https://blogs.ancestry.com/cm/whos-m...-your-sibling/

    4 sisters take a DNA test to see "who was the most Irish" results for Irish DNA according to ancestry:

    Sister 1: 45% Ireland, 32% Europe West, 16% Great Britain
    Sister 2: 54% Great Britain, 28% Ireland, 6% Scandinavia
    Sister 3: 42% Great Britain, 33% Ireland, 16% Europe West
    Sister 4: 67% Great Britain, 22% Ireland, 10% Scandinavia

    The Ireland DNA shifting from a high of 45% on the first sister to a low of 22% for the 4th is a 23 point spread, but more notable was the Great Britain category, where it runs from a high of 67% to a low of 16% a 51 point spread between siblings on those markers.

    ---------

    DNA Inheritance is random, random results are random. Yes you get ~50% of your DNA from a given Parent. Which 50% you get is the question. So going from 44% for your father, to 0% for yourself is entirely in the realm of possible, albeit not entirely probable. (But people win the Powerball Lottery Jackpot all the same)

    You also need to remember that most DNA testing sites are basing their results based on reference panels which often consist of "Person's with 'well researched family trees' going back at least 4 generations" which link them a given location. I suspect Ancestry is starting to use DNA-composits from members who aren't "pure" in their ties to certain regions as well, but think that's a more recent thing--and running directly into the jaws of the perils of what often passes for "well researched family trees."

    In other words: The tests/analysis tools give you an indication of where the DNA is(or was documented to likely be "recently"), not neccessarily where it came from. Further confusing things on your end: The Portuguese were getting around rather a LOT about five to six hundred years ago, and just about all of that is going to exist "outside of a genealogical time frame" that anyone is likely to find records for.

    Comment


    • #3
      Originally posted by bartarl260 View Post
      DNA Inheritance is random, random results are random. Yes you get ~50% of your DNA from a given Parent. Which 50% you get is the question. So going from 44% for your father, to 0% for yourself is entirely in the realm of possible, albeit not entirely probable. (But people win the Powerball Lottery Jackpot all the same)

      You also need to remember that most DNA testing sites are basing their results based on reference panels which often consist of "Person's with 'well researched family trees' going back at least 4 generations" which link them a given location. I suspect Ancestry is starting to use DNA-composits from members who aren't "pure" in their ties to certain regions as well, but think that's a more recent thing--and running directly into the jaws of the perils of what often passes for "well researched family trees."

      In other words: The tests/analysis tools give you an indication of where the DNA is(or was documented to likely be "recently"), not neccessarily where it came from. Further confusing things on your end: The Portuguese were getting around rather a LOT about five to six hundred years ago, and just about all of that is going to exist "outside of a genealogical time frame" that anyone is likely to find records for.
      Yes, I know all that. The problem is that the big difference in myOrigins between me and my father is not present in any of the others calculators: GEDmatch's Eurogenes K13, Eurogenes K15, Eurogenes K36, MDLP World MDLP World22; DNA.land; Gencove, etc...
      Besides that, I've read a lot of complaints about myOrigins 2.0 in portuguese/iberians facebook groups.

      Comment


      • #4
        Originally posted by rxavierflima View Post

        Yes, I know all that. The problem is that the big difference in myOrigins between me and my father is not present in any of the others calculators: GEDmatch's Eurogenes K13, Eurogenes K15, Eurogenes K36, MDLP World MDLP World22; DNA.land; Gencove, etc...
        Besides that, I've read a lot of complaints about myOrigins 2.0 in portuguese/iberians facebook groups.
        More importantly do you have Portuguese/Iberian matches in your match list? Your matches to real people are far more important than your ethnicity estimates and frankly more telling of the truth. Remember that the estimates about ethnicity made by these companies are simply just that, estimates.

        Comment


        • #5
          This is my family tree: https://www.dropbox.com/s/a2vxenqnl3...s%26padres.png
          Even if 1 or 2 of the red dots (persons with unknown father) are sons of non-portuguese fathers, that doesn't explain the difference between the results in myOrigins 2.0 and in other calculators.

          Comment


          • #6
            Your result appears to be an example of "inconsistency" produced by the admixture algorithms, such that a child shows a supposed ethnic component that was not present in either parent. The degree of "inconsistency" encountered in 2-parent/child trios is one of the few ways that the performance of the algorithm can be tested quantitatively on a large scale, so it is an important metric. In the documentation for a previous version of Ancestry's admixture algorithm, there was some discussion about this problem (Ancestry reported that the problem was still being encountered, but at a lower frequency, but they did not actually give any numbers). However, the documentation for Ancestry's current algorithm does not mention this issue at all! I'm not aware of any discussion of this problem by other vendors.

            It seems to me that results like this are telling us something about the inherent limitations of the methodology. They tell us that the degree of uncertainty associated with the results is much larger than we have been led to believe. The uncertainty associated with a figure of, say, 17 percent is evidently quite large, and a 0 percent result might not be statistically different from a 17 percent result. I suspect this is particularly true for European ancestry, because we know from history that large-scale migrations have taken place across the entire continent at least as far back as Roman times, to the extent that populations defined by modern national boundaries are not likely to be either homogeneous or distinct from those of other European nations.

            In addition, we have to remember that admixture algorithms use samples from modern, likely still living humans, whose ancestry is assumed to be "pure", followed by a selection process that systematically removes outliers, resulting in homogeneous "reference groups" that are then assumed to "represent" in some way the entire population of a nation or region. But from the way these "reference groups" are constructed, there is no reason to suppose that they will actually contain all, or even very much, of the genetic diversity that is or was present within any such nation or region. Admixture algorithms may be useful at the population level, but for individual genealogies, they are not dependable.

            Comment

            Working...
            X