Announcement

Collapse
No announcement yet.

Daughters match isn't a match of her parents...?

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

  • Daughters match isn't a match of her parents...?

    Hello everybody.

    I read back several pages of this subforum but i didn't find a similar thread so i created this one. Sorry if this question has already been answered before.

    My mother has a match in the family finder that shares 45cm with her with a longest block of 9,33cm.

    But none of her parents match her match. Does somebody know how this is possible? Is it a false match then?

    Kind regards, Magit

  • #2
    FTDNA declares a match if longest segment is 7cM/500SNPs or greater and Total Shared DNA is 20cM or greater. The total 20cM includes all segments down to 1cM/500 SNPs. Many segments under 7cM are false matching.

    Failing to meet the above criteria, they will also declare someone a match if they share a longest segment of 9cM or greater, no total shared DNA minimum.

    Since the longest segment is 9.3cM, This is likely to be the only true segment and its actual size is under the 9cM threshold in one of the parents, therefore FTDNA will not declare them a match. To see if it is present in one of the parents you would need to compare parents kits to this match at Gedmatch (require both your kits and matches kit to be uploaded there)
    Last edited by prairielad; 18 March 2018, 04:02 PM.

    Comment


    • #3
      Thank you for your reply!

      I checked it with the chromosome browser, and her parents basically match her everywhere just as i expected even at the spot where the match matches her.
      It is the only way to compare them as far as i know because i can't add the match to the chromosome browser of my grandparents because he is not on their matcheslist and hasn't uploaded to the gedmatch either.

      So it is not very likely that a match with a 9,33 longest matching segment is a false match?

      Comment


      • #4
        Originally posted by Magit View Post
        Hello everybody.

        I read back several pages of this subforum but i didn't find a similar thread so i created this one. Sorry if this question has already been answered before.

        My mother has a match in the family finder that shares 45cm with her with a longest block of 9,33cm.

        But none of her parents match her match. Does somebody know how this is possible? Is it a false match then?

        Kind regards, Magit
        This is a very normal experience. It's an artifact of the matching process which only indirectly distinguishes between a person's paternal and maternal chromosomes. What you are looking at is likely the 'stitching together' of a real but small inherited match (perhaps 7 cM or smaller--too small to be reflected directly with the parent) with an additional, smaller, and possibly false segment.

        The matching algorithm works by identifying contiguous blocks where at least one SNP from donor A matches at least one SNP from donor B. Because each donor has two copies of each chromosome, and there are only two possible SNP values at each location, there is something like a 94% chance that at least one SNP from donor A will match the corresponding location for donor B.

        However, along a long enough stretch of contiguous SNPs, the probability for a mis-match increases. The idea is that matches of 700 SNPs or longer are unlikely to occur by chance, and therefore most likely reflect common ancestry. This number, 700 SNPs, seems to be the industry minimum standard for reporting matches across all platforms, although some companies seem to add other restrictive criteria--like a minimum total volume--that suppresses reporting.

        I've experienced many genealogically valid matches at this minimum level, so I'm not going to complain that it is too low. But the trade-off is that the ability to detect small real matches increases the likelihood of reporting small false matches.

        If you have downloaded your raw results files to Gedmatch.com, they have a suite of enhanced analytical tools that can help you determine what the specific situation is in this case. For example, you could manually adjust downward the matching threshold in a chromosome browser comparison to this other matching donor and each of the parents. If any part of your daughter's match to the donor is valid, it should be visible in this enhanced view.

        But if you aren't immediately able to identify a likely descent scenario comparing your daughter's pedigree to this other donor, it probably isn't worth your time. Even relatively large matches--largest segment of over 20 cM--can persist for a very long time, extending well beyond the reach of most pedigrees.

        Anyhow, just for reference, I created a rough calculation to predict the % chance that a given matching segment will turn out to be false, based on the # of matching SNPs.

        500 SNPs--82%
        600 SNPs--79%
        700 SNPs--75%
        1,400 SNPs--50%
        2,800 SNPs--0%

        I have a more technical description of the calculation, but for a rough, introductory look, I think this crude summary is reasonable.

        Comment


        • #5
          Thanks a lot for your summary.

          I have already uploaded all kits i manage to Gedmatch, but my mothers' questioned match hasn't so i am not able to compare them there.

          I was interested in this match particularly because it is a prominent person in my line of work, and it would be fascinating if it realy was pointing to a legit connection.

          Comment


          • #6
            There is another strategy you may be able to employ at Gedmatch while you are waiting for this other person to maybe download their kit.

            You could subscribe to Gedmatch's premium level service, Tier One, and use the Matching Segment Analysis tool to identify all the people that match a given kit on a specific chromosome. If anyone matches the parents along this region of the chromosome in question, you can try to establish a line of descent that could connect all three of you.

            Tier One is only $10 a month, and you can sign up for a single month if you want to. There is a risk that nobody will match your parents in this region, but it's still a good investment to map out your known matches in this way to streamline the process of identifying common ancestors.

            Comment


            • #7
              Come to think of it, you can try something similar at Family Finder by downloading the chromosome browser details for all your matches into an Excel file and filtering to focus on the region you are interested in.

              Actually, you'll probably want to try both strategies because, as you pointed out, not everyone at Family Finder has a kit at Gedmatch, and vice versa.

              Comment


              • #8
                @ Frederator
                Thanks a lot, that's a good idea!
                I guess i will try that in my next vacations.

                Comment


                • #9
                  This often happens to me. I will have a small segment that neither parent has. My father was Italian and my mother has no Italian ancestry, so I can just look at the In Common list and see whose side the match is on.

                  Another thing I find is that although I don't share the segment with my mother, often I do share it with her 1st cousin.

                  Comment


                  • #10
                    That situation sounds like a data problem (data errors or no-calls?), or, perhaps, a combination of a larger maternal with a smaller paternal segment that is just enough to trigger the matching algorithm.

                    Most of the time, matching segments are well behaved and will fall into clean triangulation groups. However, I have a few segments where some kits match in all combinations (triangulation), while other kits match only some of those, and not in all combinations. To me, this suggests that either the data or the properties of individuals SNP's have caused the matching algorithms to misbehave, yielding misleading results. When illogical results occur, it seems best not to base genealogical conclusions on them!

                    Comment


                    • #11
                      Thanks to all of you for your replies.
                      I will consider it a "maybe" - "maybe not" connection.

                      Comment


                      • #12
                        Originally posted by Frederator View Post
                        This is a very normal experience. It's an artifact of the matching process which only indirectly distinguishes between a person's paternal and maternal chromosomes. What you are looking at is likely the 'stitching together' of a real but small inherited match (perhaps 7 cM or smaller--too small to be reflected directly with the parent) with an additional, smaller, and possibly false segment.

                        The matching algorithm works by identifying contiguous blocks where at least one SNP from donor A matches at least one SNP from donor B. Because each donor has two copies of each chromosome, and there are only two possible SNP values at each location, there is something like a 94% chance that at least one SNP from donor A will match the corresponding location for donor B.

                        However, along a long enough stretch of contiguous SNPs, the probability for a mis-match increases. The idea is that matches of 700 SNPs or longer are unlikely to occur by chance, and therefore most likely reflect common ancestry. This number, 700 SNPs, seems to be the industry minimum standard for reporting matches across all platforms, although some companies seem to add other restrictive criteria--like a minimum total volume--that suppresses reporting.
                        This "stitching together" a false match is a supposition frequently made, but never really proven. From what I've seen is that almost inevitably a true reconstruction is made of a segment of the common ancestor's DNA.

                        You talk about a 94% chance that a donor at a particular SNP will get at least a half match with the other donor being compared. I think that chance is actually a bit smaller, but using your 94%, the chances of two unrelated people matching one hundred SNP's in a row is 0.94 to the 100 power, or a chance of about 0.002, one out of five hundred.

                        If two completely unrelated people have almost no chance of matching on just 100 SNP's, why would the stitching together of the DNA of two people who share a common ancestor produce anything else, but a match to that common ancestor.

                        So all of these short matching segments are not false positives, though many are not IBD (Identical by Distant) but reconstructed segments. However, some experts make statements about large matching segments from distant relatives saying that segments of at least 10 cM in length are undoubtedly IBD (and the bigger, the better) . Actually from the viewpoint of logic and probability, that result is a near impossibility. Clearly a small segment is more likely to be IBD than a large segment (for distant matches).

                        Once you go more than a few generations back, the current matching methodology is completely bankrupt. A reconsideration is long overdue.

                        Jack Wyatt
                        Last edited by georgian1950; 21 March 2018, 12:54 PM. Reason: syntax

                        Comment


                        • #13
                          Originally posted by georgian1950 View Post
                          This "stitching together" a false match is a supposition frequently made, but never really proven. From what I've seen is that almost inevitably a true reconstruction is made of a segment of the common ancestor's DNA.

                          You talk about a 94% chance that a donor at a particular SNP will get at least a half match with the other donor being compared. I think that chance is actually a bit smaller, but using your 94%, the chances of two unrelated people matching one hundred SNP's in a row is 0.94 to the 100 power, or a chance of about 0.002, one out of five hundred.

                          If two completely unrelated people have almost no chance of matching on just 100 SNP's, why would the stitching together of the DNA of two people who share a common ancestor produce anything else, but a match to that common ancestor.

                          So all of these short matching segments are not false positives, though many are not IBD (Identical by Distant) but reconstructed segments. However, some experts make statements about large matching segments from distant relatives saying that segments of at least 10 cM in length are undoubtedly IBD (and the bigger, the better) . Actually from the viewpoint of logic and probability, that result is a near impossibility. Clearly a small segment is more likely to be IBD than a large segment (for distant matches).

                          Once you go more than a few generations back, the current matching methodology is completely bankrupt. A reconsideration is long overdue.

                          Jack Wyatt
                          I've done the actual math, and 94% is the correct number for the probability of matching at least at the half-identical level at any single SNP.

                          Using Excel's binomial distribution assuming:

                          A.) 0 successes
                          B.) 4 trials (i.e., two chromosome copies for each of 2 donors)
                          C.) 1/2 probability (i.e., only 2 possible values for any given location)

                          The result is roughly 6%. 100% less roughly 6%=roughly 94%, the probability of at least a half-identical match for a single SNP. You're not going to get any number other than this.

                          From there the probability of any number of consecutive SNPs matching at least the half-identical level is pretty simple.
                          Your example:

                          A.) 100 successes
                          B.) 100 trials
                          C.) 94% probability

                          The actual odds are 1 in 635.14.

                          In other words, two (2) randomly selected individuals out of a pool of about thirty-seven (37) people are guaranteed to match on at least 100 consecutive SNPs, whether they're IBD or IBS. Thirty-seven people (37) can people can be grouped into 666 pairs of two individuals.

                          To be blunt: A match of only 100 SNPs is meaningless.

                          My own downstream calculations are based on the industry standard assumption of potential statistical significance only above 700 consecutive matching SNPs, which is well established. If you don't like that, take it up with ISOGG.

                          But I guarantee you will be wasting your time. You would be swimming against the current of the hundreds and thousands of published accounts of matches at the 700 SNP level falling apart upon the phasing of even one kit, not to mention both kits. You would have to establish the empircal superiority of your lower threshold in the context of an already unreliable higher threshold.

                          Comment


                          • #14
                            Originally posted by Frederator View Post
                            But I guarantee you will be wasting your time. You would be swimming against the current of the hundreds and thousands of published accounts of matches at the 700 SNP level falling apart upon the phasing of even one kit, not to mention both kits. You would have to establish the empircal superiority of your lower threshold in the context of an already unreliable higher threshold.
                            We will get back to the rest later because you have misconceptions and errors, but let's talk about phasing.

                            We are discussing a situation where both parents have DNA from a common ancestor which combine to reconstruct a segment of DNA from the common ancestor. Which parent's DNA are you throwing out? Throw out either one and of course matches are going to fall apart. Phasing is an absurd part of the theory except maybe for matches within the first few generations.

                            Genetic genealogy commits what I call the chain letter fallacy. Chain letters work, provided the population is infinite and each round you are able to find people who have not played before.

                            In reality, we have a hell of a lot of common ancestry, most of which is with the genealogical time frame.

                            Jack Wyatt

                            Comment


                            • #15
                              Originally posted by georgian1950 View Post
                              We will get back to the rest later because you have misconceptions and errors, but let's talk about phasing... .
                              No we won't because you're cut off. I've shown my work and all you've done is flap your gums.

                              You're finished.

                              Comment

                              Working...
                              X