False-Positives in Autosomal DNA Matches

6

I have spent a significant amount of time and effort beating my head on brickwalls that appear when tracking Ancestry Autosomal DNA matches. I've had success in establishing document trails with matches as low as 7 CM and no success with other matches in the 20-25 cM range.

So I'm wondering just how much time (and bloody forehead) is warranted when trying to establish a actual cousin relationship at lower cM match levels.

I think that I already know that a certain percentage (an average) of actual cousin relationships will not be detected by autosomal DNA. See ISOGG Cousins statistics These I would call false-negatives.

What I don't know about is any study that give clues as to the percentage of detected autosomal DNA matches that are not an actual cousin relationship. I would refer to these as false-positives or error matches.

Consider the following: (all are Ancestry Autosomal DNA matches)

  Helen and Nancy have a 11   cM match
Helen and Sally Have a 21   cM match
Nancy and Sally have a 1900 cM match


What is the likeihood that both (simultanously) the Helen/Sally and the Helen/Nancy matches are false-positives?

(By the way this is not theoretical, but rather an actual example in my family)

Are those cM values the match to each other, or to you? In other words, are you only looking from your own results, or do you have at least Viewer access to the results of Helen, Nancy and Sally? – PolyGeo – 2020-01-13T05:13:40.397

Have you checked for pile-ups? – ColeValleyGirl – 2020-01-13T08:04:46.333

Re your failures to identify an actual cousin relationship. What classification would you give to a non paternal event, where the biological father is not the paper based father? NPEs would presumably result in a failure to find the link on paper. Estimates of NPEs seem to vary wildly - Wikipedia has stuff on the topic with some researchers suggesting 2% and describing figures like 30% as urban myths. Regardless, I presume that NPEs will account for some of your DNA matches that don't reveal matches on paper. – AdrianB38 – 2020-01-13T13:54:29.937

@PolyGeo I happen to manage Nancy's DNA, so Nancy's DNA match with Helen and Sally are results of direct reporting of Nancy's DNA matches (as distinguished from Ancestry's shared match serach).With regard to Helen, I am corresponding directly with her, and she reported that the simple DNA match (that is, not the "shared" match). None of these cM relate to me. – BobE – 2020-01-13T15:15:02.837

@ColeValleyGirl - I've been "told" that Ancestry takes extra effort to exclude pile-ups, and more specifically because Ancestry uses something called "phasing" the accuracy of their DNA matches is improved (but obviously not foolproof). So the short answer to your question is "no", and I wouldn't even know where to begin. – BobE – 2020-01-13T15:39:23.143

@Adrian - I'm assuming that you are referring to false or miss-attributed paternity (or what is somewhat crassly called bastardy). While NPE should not be ruled out, when the suspects of NPE are in wildly different regions of the world with no evidence of contact NPEs become a less viable alternative explanation. (I would be suspicious of a suggestion that small plot Turkish farmer in the early 1800s impregnated a farmer's daughter in Iowa when there is no evidence that either traveled much more than beyond their respective counties borders.) – BobE – 2020-01-13T16:00:50.510

That is what I'm referring to, Bob. But your interpretation of who begat who on whom confuses me because surely shared cM can't identify putative parents??? – AdrianB38 – 2020-01-13T16:15:43.010

@Adrian - then perhaps I'm not understanding your original comment. Perhaps it would be best to discuss this elsewhere as I don't see the point you are making (as it applies to this question) – BobE – 2020-01-13T16:44:03.100

5

Ancestry DNA uses scientific methods to attempt to give you valid matches. They use their own proprietary technique to detect Identical by Descent (IBD) segments and to match people. They work to ensure as few false-positives as possible. Much of their methodology is described in their Ancestry DNA Matching White Paper.

Based on Ancestry's methodology, I would think that the Helen/Sally and the Helen/Nancy matches are probably valid. I would extend that to say the vast majority of your Ancestry matches are most likely valid and are truly DNA related to you. But finding the genealogical relationship of that DNA path is difficult, if not impossible in many cases. There's a few reasons:

1. If you do find a genealogical relationship with a DNA match, that does not necessarily mean that you've found the correct DNA path. The common ancestor who passed down the DNA to the two of you may not be the genealogical common ancestor you found.
2. You may have multiple DNA relationships via multiple matching segments. Just because you have found one genealogical relationship with a DNA match, that doesn't mean it is the only one if you have more than one segment match.
3. Segments can be 10 or more generations old. Can you genealogically trace all your lines and all your DNA matches lines back at least 10 generations? Any one line you don't have at least 10 generations back might end up being that DNA connection.
4. And of course there are the genealogical documents that say one thing, but the true father and/or mother of the DNA line is someone else. With an estimated 1 to 2 percent chance of misattributed parentage, many DNA lines become unidentifiable.

There are many techniques available that can help you narrow down the possible ancestral lines of your DNA matches. At Ancestry DNA, you can use The Leeds Method, automated clustering such as Genetic Affairs or Ancestry's own hybrid DNA/tree hinting system called ThruLines. You can then get sophisticated and combine them all as Jim Bartlett explains in his Walking The Clusters Back series. You can also upload your raw data to other sites such as GEDmatch that will allow you to analyze individual segments and build up triangulation groups. Working with all your matches together gives you slightly better odds on finding some of your DNA matches' relationships than does working on them alone, one by one. These techniques will allow to you find a few and maybe some of your DNA match relationships.

A DNA connection is like a clue to genealogists. It gives you a person who you might be able to add to your tree. If you find a connection, that's great. If you don't, it still might lead you to finding other documents and possible relatives that could extend your genealogical research on some of your branches. If your DNA matches are doing that for you, then be happy that they are serving you well.