Ann, FTDNA has posted the probabilities of a specific relative passing their filter and showing up as a match with us. Parents and 1st cousins - virtually 100 percent 2nd cousins about 99 percent 3rd cousins about 90 percent 4th cousins about 50 percent declining percentages beyond that And so some people literally throw in the towel at 5th cousin and beyond. But that's wrong! These are percentages of a specific person that we might select for testing. And it's true that if we get a 5th cousin to test, he/she probably won't show up as a match. BUT, we have many, many 5th cousins; and when you multiply the large number of 5th cousins times the small percentage that a specific one will match, the odds are higher that you'll get a match with a 5th cousin that with a 4th cousin. You'll actually get more 6th cousins than 5th - I think it probably levels off at about 7th or 8th cousins. Think about how many matches you have, and the fact that you and your match would probably recognize a 3rd cousinship. So those matches must be more distant. Look at it another way. Look at the number of shared cM for a close match that you do have. How many of your matches share that amount of atDNA with you - very few. Most of our matches are in the 8-12cM range. Well at that range they are probably 7th cousins or so. If you believe the 1-in-a-million type of statement, then you must be thinking that virtually none of the matches reported to your are related - that's silly. Most of them are related, we just can find the relationship. IMO Jim Bartlett On 05/15/12, Ann Turner<[email protected]> wrote: [1]http://genealogy.about.com/b/2012/05/15/autosomal-dna-testing.htm?nl=1 I want to post something sometime in that very interesting long thread about GEDCOM files, but this article mentions finding close connections at all three autosomal testing services. Quoting Kathy Johnston from that thread "5. Even if I go back 4 generations, am I likely to randomly match any true 3rd cousins? NO!! because you need a database of over a million to find that one in a million 3rd cousin. Even if you have more than 300 3rd cousins alive today, there are over 300 million people living in the U.S. Does FTDNA test that many people? No." I have got to think more about the statistics here. Obviously we are going to hear a disproportionate number of success stories, just because they're so much fun. But it strikes me that we are hearing too many stories for the one-in-a-million number to hold true. I have a 2nd cousin once removed at 23andMe who tested because his daughter worked at Illumina and got a free kit. I tested a known cousin at FTDNA, and when she got her results back, she recognized the name of a 3rd cousin once removed from another side of her family. I do think FTDNA overcalls the 3rd cousin bin, relying too much on the length of the longest segments. Confirmed 3rd cousins tend to have multiple segments over 5 cM in length. But why are we seeing so many stories about confirmed close cousins? Just thinking out loud here -- maybe this is related to the "birthday paradox": [2]http://en.wikipedia.org/wiki/Birthday_problem Ann Turner References 1. http://genealogy.about.com/b/2012/05/15/autosomal-dna-testing.htm?nl=1 2. http://en.wikipedia.org/wiki/Birthday_problem

Jim, I think you (and Debbie, too) are expanding the scope of my message beyond what I intended. It was specifically limited to 3rd cousins and Kathy Johnston's statement that the odds of finding a *random* match as close as a 3rd cousin were one-in-a-million. My response was that it seems like we are seeing too many anecdotal reports of success for that number to make sense, but I need to think more about the statistics. I don't disagree with you about the large number of distant cousins, but we need to be extremely cautious about attributing the matching segment to a specific common ancestor. Remember how you and my nephew had a matching segment based on genotype data? It triggered my search through your genealogy file at RootsWeb, where your paper trail broke through a brick wall for me. Later I realized that the segment was not present in my sister or her husband, a red flag for Identity by State. In fact, it turned out to be a pseudo-segment when I looked at haplotype data. Ann On Tue, May 15, 2012 at 3:45 PM, Jim Bartlett <[email protected]>wrote: > > Ann, > > FTDNA has posted the probabilities of a specific relative passing their > filter and showing up as a match with us. > > Parents and 1st cousins - virtually 100 percent > 2nd cousins about 99 percent > 3rd cousins about 90 percent > 4th cousins about 50 percent > declining percentages beyond that > > And so some people literally throw in the towel at 5th cousin and beyond. > But that's wrong! These are percentages of a specific person that we > might > select for testing. And it's true that if we get a 5th cousin to test, > he/she probably won't show up as a match. > > BUT, we have many, many 5th cousins; and when you multiply the large > number > of 5th cousins times the small percentage that a specific one will match, > the odds are higher that you'll get a match with a 5th cousin that with a > 4th cousin. You'll actually get more 6th cousins than 5th - I think it > probably levels off at about 7th or 8th cousins. > > Think about how many matches you have, and the fact that you and your > match > would probably recognize a 3rd cousinship. So those matches must be more > distant. > > Look at it another way. Look at the number of shared cM for a close match > that you do have. How many of your matches share that amount of atDNA > with > you - very few. Most of our matches are in the 8-12cM range. Well at that > range they are probably 7th cousins or so. > > If you believe the 1-in-a-million type of statement, then you must be > thinking that virtually none of the matches reported to your are related > - > that's silly. Most of them are related, we just can find the > relationship. > > IMO > > Jim Bartlett > > On 05/15/12, Ann Turner<[email protected]> wrote: > > [1] > http://genealogy.about.com/b/2012/05/15/autosomal-dna-testing.htm?nl=1 > I want to post something sometime in that very interesting long thread > about GEDCOM files, but this article mentions finding close connections > at > all three autosomal testing services. Quoting Kathy Johnston from that > thread > "5. Even if I go back 4 generations, am I likely to randomly match any > true > 3rd cousins? NO!! because you need a database of over a million to find > that one in a million 3rd cousin. Even if you have more than 300 3rd > cousins alive today, there are over 300 million people living in the U.S. > Does FTDNA test that many people? No." > I have got to think more about the statistics here. Obviously we are > going > to hear a disproportionate number of success stories, just because > they're > so much fun. But it strikes me that we are hearing too many stories for > the > one-in-a-million number to hold true. I have a 2nd cousin once removed at > 23andMe who tested because his daughter worked at Illumina and got a free > kit. I tested a known cousin at FTDNA, and when she got her results back, > she recognized the name of a 3rd cousin once removed from another side of > her family. > I do think FTDNA overcalls the 3rd cousin bin, relying too much on the > length of the longest segments. Confirmed 3rd cousins tend to have > multiple > segments over 5 cM in length. But why are we seeing so many stories about > confirmed close cousins? Just thinking out loud here -- maybe this is > related to the "birthday paradox": > [2]http://en.wikipedia.org/wiki/Birthday_problem > Ann Turner > > References > > 1. > http://genealogy.about.com/b/2012/05/15/autosomal-dna-testing.htm?nl=1 > 2. http://en.wikipedia.org/wiki/Birthday_problem > > > ______________________________ > For answers to Frequently Asked Questions about mailing lists, please see: > http://dgmweb.net/MailingListFAQs.html > > > ------------------------------- > To unsubscribe from the list, please send an email to > [email protected] with the word 'unsubscribe' without > the quotes in the subject and the body of the message >