Announcement

Collapse
No announcement yet.

novice mistrusts FTDNA. any comments?

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

  • novice mistrusts FTDNA. any comments?

    I will go ahead with the disclaimers first:
    I am just a lay person.. I have very little knowledge of DNA testing, genetics and biology. Additionally I am not the project leader of my surname project.


    Recently I asked an FTDNA rep the following question:

    Suppose 2 subjects match 37/37. FTDNA's chart tell us that there's about a 95% chance that an MRCA occurs by 7 generations. Suppose these 2 subjects could verify that there were no MRCA in 8 generations in each of their genealogies. Should we assume

    a) that there is a greater than 95% that an MRCA occurs in the 9th generation; or in other words, there is a greater than 95% chance that the earliest known ancestor of each of the two subjects were brothers

    b) that there is a 95% chance that the MRCA occurs by the 15th generation
    (8 generations with no possible MRCA + 7 generations of potential MCRAs)

    c) something other than a or b; Please explain.
    -----

    The FTDNA reps' reply was prompt and courteous but, I suspect, not 100% honest.

    The rep said there were many things to consider, among these factors were surname and haplogoup. The explanation included considering whether or not the two subject were member of the same haplogroup.....

    but I thought the haplogroups were dependant on the test values! would not a 37/37 match, by definition, place the two subjects in the same haplogroup???....

    The rep went on for several paragraphs never mentioning how anything related to odds, probabilities or generation ranges in the context of my question. Perhaps I am jumping to conclusions here, but I took that as intentional attempt to confuse a common lay person and avoid the question he asked...........
    ---

    I didn't mention this to the ftdna rep, but in my surname project, 7 other members match me 37/37 yet apparently none of us have a documented common ancestor. FTDNA's is pretty optimistic about 37/37 matches, a chart sent to my by my group leader says that there is 59% chance that MRCA occurs in just *2* generations!

    I am not sure how much genealogical info my group leader has on the 8 of us that match 37/37 but I would bet he would have more than 2 generation on each of us. There should be (7+6+5+4+3+2+1 = 28) possible parings of us 8 yet none of us appear to have a documented common ancestor......BUT FTDNA tells us that there is a 59% chance of a MCRA in just 2 generations!
    Imagine that you flip a coin 28 times, each flip has 50% chance of landing on heads or tails. How likely do you think that in each of 28 flips the coin lands on tails? I can think of only 5 possible reasons for the no documented common ancestors in our group of 8 that match 37/37 with same surname :

    a) very very very unlikely statistical event
    b) basic flaw in my logic or my understanding of problem
    c) failure by project leader to identify common relatives in our group of 8's genealogies: I am pretty certain that each has at least 2 generations of documented paternal ancestors but I suppose it is possible the group leader has overlooked a documented match in the group of 8
    d) some type of failure in testing procedure.... (perhaps lab assistant needed to study for midterm and instead of running test, merely copied results from previous tests of our surname)
    e) FTDNA's statistics are totally bogus.

    In light of the FTDNA rep's apparent attempt to confuse me instead of answering a simple question, I am inclined to distrust FTDNA and think that the cause is most likely D or E. Can anyone explain to me why I am wrong or have you had a smilar experience with FTDNA?
    Bertp
    FTDNA Customer
    Last edited by Bertp; 22 September 2006, 08:01 PM.

  • #2
    Very Interesting

    I find your post very interesting. I think one needs to look at qualifiers in any of the promises made by any company. I would hesitate to charge anyone deliberate intent to deceive, but is it ethical to promise to find out, say, Jewish ancestors unless one is speaking of descendants of kohanim?
    I was tested by FTDNA and got very prompt replies to my questions. I havre as many doubts as answers. I had intended to write an article about my experience, but my general feeling at the end was so negative that I decided to bag the project. I am not a scientist, but scientists with whom I talked frowned at the promises made by FTDNA. Maybe FTDNA itself should clear up in unequivocal language whar it can and cannot do. Personally, I am very sceptical of many of its claims. I have chosen to confine my comments to this forum. Comments, Bennett?

    Comment


    • #3
      Bertp,

      You may be a newbie but you be a newbie with a noggin! Rock on pussycat!

      Fortunately, for you, and all the other reality-oriented testees, FTDNA has an answer - the 67-marker test!

      MRCA? MZRCA? It'll drive one BZRCA!!

      Tom

      PS I wish I was in your surname group (or you in mine). Lead. I will follow.

      Comment


      • #4
        Bertp,

        This is an astonishing situation! What is the url for this surname project's website? I'd like to look at the haplotypes.

        Mike Maddi

        Comment


        • #5
          What is so astonishing about this? 95% does not equal 100%. There have to be exceptions and situations that fall into that remaining 5%. Please remember these are just probabilities, they are not guarantees. I'm sure FTDNA is not purposely trying to confuse or mislead anyone.

          I have exact 37-marker matches in my surname project as well, but unfortunately do not have the ability to even go back 8 generations in our genealogies, due to lack of records available or easily obtainable for this region of Eastern Europe. So for now I have to settle with just knowing that the families are indeed related and it will probably be many years before we figure out who the common ancestors were -- if ever! People with genealogies going back to the 1600-1700s are already way ahead of the game, as far as I'm concerned.
          efgen
          Webinar Coordinator
          Last edited by efgen; 23 September 2006, 01:03 PM.

          Comment


          • #6
            Mike, here is the website: http://www.familytreedna.com/(ela3xm55n40utd550usfjymg)/public/PittmanDNA/index.aspx

            It hasn't been updated since the last results were received, but new member #34 belongs to the group of 8 that I mentioned

            By the way, while trying to learn about haplogroups, I found a thread that mentioned my surname. however I was unable to make heads or tails of it.
            http://searches2.rootsweb.com/cgi-bi...+35157721606+F

            Comment


            • #7
              efgen,
              With all due respect, I was not disputing the 95% after 7 generations probability. In fact, I wrote to the ftdna rep only after a discussion with another group member who I suspected was making the too optimistic assumptions about the specific British origins of our line. Its been many years since I sat in a math class but, I felt b) was the correct answer to the question I posed to the FTDNA rep and I was hoping the rep could confirm which , if either ;-) was correct.
              Bertp
              FTDNA Customer
              Last edited by Bertp; 23 September 2006, 01:59 PM.

              Comment


              • #8
                Originally posted by Bertp
                I will go ahead with the disclaimers first:
                I am just a lay person.. I have very little knowledge of DNA testing, genetics and biology. Additionally I am not the project leader of my surname project.


                Recently I asked an FTDNA rep the following question:

                Suppose 2 subjects match 37/37. FTDNA's chart tell us that there's about a 95% chance that an MRCA occurs by 7 generations. Suppose these 2 subjects could verify that there were no MRCA in 8 generations in each of their genealogies. Should we assume

                a) that there is a greater than 95% that an MRCA occurs in the 9th generation; or in other words, there is a greater than 95% chance that the earliest known ancestor of each of the two subjects were brothers

                b) that there is a 95% chance that the MRCA occurs by the 15th generation
                (8 generations with no possible MRCA + 7 generations of potential MCRAs)

                c) something other than a or b; Please explain.
                -----

                The FTDNA reps' reply was prompt and courteous but, I suspect, not 100% honest.

                The rep said there were many things to consider, among these factors were surname and haplogoup. The explanation included considering whether or not the two subject were member of the same haplogroup.....

                but I thought the haplogroups were dependant on the test values! would not a 37/37 match, by definition, place the two subjects in the same haplogroup???....

                The rep went on for several paragraphs never mentioning how anything related to odds, probabilities or generation ranges in the context of my question. Perhaps I am jumping to conclusions here, but I took that as intentional attempt to confuse a common lay person and avoid the question he asked...........
                ---

                I didn't mention this to the ftdna rep, but in my surname project, 7 other members match me 37/37 yet apparently none of us have a documented common ancestor. FTDNA's is pretty optimistic about 37/37 matches, a chart sent to my by my group leader says that there is 59% chance that MRCA occurs in just *2* generations!

                I am not sure how much genealogical info my group leader has on the 8 of us that match 37/37 but I would bet he would have more than 2 generation on each of us. There should be (7+6+5+4+3+2+1 = 28) possible parings of us 8 yet none of us appear to have a documented common ancestor......BUT FTDNA tells us that there is a 59% chance of a MCRA in just 2 generations!
                Imagine that you flip a coin 28 times, each flip has 50% chance of landing on heads or tails. How likely do you think that in each of 28 flips the coin lands on tails? I can think of only 5 possible reasons for the no documented common ancestors in our group of 8 that match 37/37 with same surname :

                a) very very very unlikely statistical event
                b) basic flaw in my logic or my understanding of problem
                c) failure by project leader to identify common relatives in our group of 8's genealogies: I am pretty certain that each has at least 2 generations of documented paternal ancestors but I suppose it is possible the group leader has overlooked a documented match in the group of 8
                d) some type of failure in testing procedure.... (perhaps lab assistant needed to study for midterm and instead of running test, merely copied results from previous tests of our surname)
                e) FTDNA's statistics are totally bogus.

                In light of the FTDNA rep's apparent attempt to confuse me instead of answering a simple question, I am inclined to distrust FTDNA and think that the cause is most likely D or E. Can anyone explain to me why I am wrong or have you had a smilar experience with FTDNA?
                Perhaps I misunderstood your original posting, above. I thought you were saying that there were 8 people with the same surname, Pitman/Pittman in this case, who matched 37/37 and none of whom could find a common ancestor in their family trees in the last 8 generations. That's what I found astonishing, since that situation would seem to be impossible, given mutation rates and the laws of statistical probability.

                Re-reading your post, I think you are saying that only you and one of the other project members with matching 37 marker haplotypes have checked your family trees and haven't found an MRCA in the last 8 generations. That is not so astonishing.

                FTDNA's estimation of a 95% probability that the two of you share an MRCA in the last 7 generations includes a 5% probability that your MRCA is further back than 7 generations. That seems unlikely, when you hear 95%, but that's certainly possible when you combine mutation rates with statistical probability. It merely means that a small percent of the population will have a slightly different mutation rate that puts an MRCA a little further back.

                So I see nothing wrong in this situation in terms of what FTDNA has predicted. The point is that you're in a surname project and the expectation is that project members are going to match. The fact that so far none of the 8 of you with the same haplotype has yet found the MCRA is not necessarily defying the probabilities that FTDNA states. You mention that you and the other person each have family trees that go back 8 generations, but you don't give a figure for the family trees of the other 6 people. Having researched the records for my family tree for 5 years, I know personally that mistakes occur, either because the record keeper made a mistake in someone's name or because I mistook a person with the same exact given name of my ancestor as my ancestor, when the person shared the same exact name by coincidence. So there is a possibility that the records may be wrong and that is one reason that genetic genealogy can help in paper trail research.

                I'm sorry that the FTDNA representative did not answer your concerns as well as he or she might have. However, I don't think it's warranted to claim that FTDNA is trying to "dupe" you in this case.

                Mike

                Comment


                • #9
                  I did a little test to see if it's possible that your haplotype is so common among R1b's that it will easily match many R1b's who have no relation to you at all. Everybody has heard of people who have tested and are R1b having hundreds of close matches at the 12 marker level. This is because their marker values are so common.

                  At 37 markers, there are enough mutations possible that chance will not allow two people who are not related in the last few hundred years to share a close match, say 34 or more markers.

                  So I went to ysearch and used only 9 markers from your haplotype to search for matches. I used a combination of markers that would be very common among R1b's and a few that are uncommon among R1b's. Here are the markers and their values that I used:

                  393=12
                  390=23
                  19=14
                  391=11
                  426=12
                  388=12
                  439=12
                  CDYa=36
                  CDYb=36

                  Here's the link for the 10 people in ysearch who match that 9 marker haplotype exactly - http://tinyurl.com/hwrya All 10 are named Pitman or Pittman. You would think just using a 9 marker R1b haplotype, that some other surnames would show up as matches. The fact that only Pitman/Pittman's show up on the list convinces me that there definitely is a relationship among these 10 in the last few hundred years.

                  Mike

                  Comment


                  • #10
                    Sorry guys, apparently I overestimated my ability to write clearly.

                    The first post in this thread was divided into two parts:

                    a) The first part was a simple probability question. I was perturbed and suspicious that the FTDNA would not give me a clear cut answer to a simple math question that they must get often. I included it here because I thought that if FTDNA would not answer then maybe I could get a consensus opinion here.

                    b) as I reread the second part, I still don't see how it was misunderstood. I was talking about 8 37/37 matches where at least 2 generations for each subject can be checked . As I mentioned, FTDNA rates the probability of a MCRA in 2 generations as 59%. To me, the analogy about the 28 coin tosses was convincing.. I would even characterize it as an "astonishing situation" that one can flip a coin 28 times and it lands on tails every single time. Unless I am making a big error in reasoning, the fact that none of our 28 parings has a MRCA in **2** generations, should be even less likely than 28 of 28 flips of a coin landing on tails.

                    However, despite my difficulty in stating my thoughts clearly and concisely, I do sincerely appreciate your comments
                    Bertp
                    FTDNA Customer
                    Last edited by Bertp; 23 September 2006, 07:48 PM.

                    Comment


                    • #11
                      Originally posted by Bertp
                      b) as I reread the second part, I still don't see how it was misunderstood. I was talking about 8 37/37 matches where at least 2 generations for each subject can be checked . As I mentioned, FTDNA rates the probability of a MCRA in 2 generations as 59%. To me, the analogy about the 28 coin tosses was convincing.. I would even characterize it as an "astonishing situation" that one can flip a coin 28 times and it lands on tails every single time. Unless I am making a big error in reasoning, the fact that none of our 28 parings has a MRCA in **2** generations, should be even less likely than 28 of 28 flips of a coin landing on tails.
                      Here's the error in your reasoning about making an analogy of the coin toss to the situation of the 8 37/37 matches in a surname project. Each coin toss is totally unrelated to the previous x number of coin tosses and the next x number of coin tosses. That is not analagous to a situation where 8 men match exactly at 37 markers and have the same surname. If there is a possibility that the 8 men share just one paternal line, then you have to take into account that there is possibly only one "coin toss" among the 8 of them. And if they are indeed in the same paternal line, then each will very likely have the same 37 marker haplotype - one "coin toss" with the same result in each of the related men. In other words, the "coin toss" in this case is not 8 men, but the paternal line involved. If there were 8 different paternal lines involved in this comparison, then your analogy would hold true, otherwise not. That is the whole basis of DNA testing being applied to genealogy.

                      If the formula that FTDNA uses states that there is a 95% probability that the 8 have a MRCA in the last 7 generations, there is still a 5% chance that the MRCA is beyond 7 generations, let's say 10 generations. The formula is computed taking an average mutation rate observed in various studies and projects. As an average, there are many around the average mutation rate and as you move further from the average, there are much less at either the slow mutating and fast mutating end of the spectrum. What we are probably seeing in your paternal line is slower than normal mutation, so you have to look back more generations than in 95% of 37/37 matches.

                      That was a bit meandering. Does it make sense?

                      Mike

                      Comment


                      • #12
                        Originally posted by Bertp
                        I will go ahead with the disclaimers first:
                        I am just a lay person.. I have very little knowledge of DNA testing, genetics and biology. Additionally I am not the project leader of my surname project.


                        Recently I asked an FTDNA rep the following question:

                        Suppose 2 subjects match 37/37. FTDNA's chart tell us that there's about a 95% chance that an MRCA occurs by 7 generations. Suppose these 2 subjects could verify that there were no MRCA in 8 generations in each of their genealogies. Should we assume

                        a) that there is a greater than 95% that an MRCA occurs in the 9th generation; or in other words, there is a greater than 95% chance that the earliest known ancestor of each of the two subjects were brothers

                        b) that there is a 95% chance that the MRCA occurs by the 15th generation
                        (8 generations with no possible MRCA + 7 generations of potential MCRAs)

                        c) something other than a or b; Please explain.
                        -----

                        The FTDNA reps' reply was prompt and courteous but, I suspect, not 100% honest.

                        The rep said there were many things to consider, among these factors were surname and haplogoup. The explanation included considering whether or not the two subject were member of the same haplogroup.....

                        but I thought the haplogroups were dependant on the test values! would not a 37/37 match, by definition, place the two subjects in the same haplogroup???....

                        The rep went on for several paragraphs never mentioning how anything related to odds, probabilities or generation ranges in the context of my question. Perhaps I am jumping to conclusions here, but I took that as intentional attempt to confuse a common lay person and avoid the question he asked...........
                        ---

                        I didn't mention this to the ftdna rep, but in my surname project, 7 other members match me 37/37 yet apparently none of us have a documented common ancestor. FTDNA's is pretty optimistic about 37/37 matches, a chart sent to my by my group leader says that there is 59% chance that MRCA occurs in just *2* generations!

                        I am not sure how much genealogical info my group leader has on the 8 of us that match 37/37 but I would bet he would have more than 2 generation on each of us. There should be (7+6+5+4+3+2+1 = 28) possible parings of us 8 yet none of us appear to have a documented common ancestor......BUT FTDNA tells us that there is a 59% chance of a MCRA in just 2 generations!
                        Imagine that you flip a coin 28 times, each flip has 50% chance of landing on heads or tails. How likely do you think that in each of 28 flips the coin lands on tails? I can think of only 5 possible reasons for the no documented common ancestors in our group of 8 that match 37/37 with same surname :

                        a) very very very unlikely statistical event
                        b) basic flaw in my logic or my understanding of problem
                        c) failure by project leader to identify common relatives in our group of 8's genealogies: I am pretty certain that each has at least 2 generations of documented paternal ancestors but I suppose it is possible the group leader has overlooked a documented match in the group of 8
                        d) some type of failure in testing procedure.... (perhaps lab assistant needed to study for midterm and instead of running test, merely copied results from previous tests of our surname)
                        e) FTDNA's statistics are totally bogus.

                        In light of the FTDNA rep's apparent attempt to confuse me instead of answering a simple question, I am inclined to distrust FTDNA and think that the cause is most likely D or E. Can anyone explain to me why I am wrong or have you had a smilar experience with FTDNA?
                        they past all possible tests i have seen
                        and i trust them more then some unknown geaneologistal research in another country who charges more money for something which is doubious at best

                        Comment


                        • #13
                          Originally posted by MMaddi
                          Here's the error in your reasoning about making an analogy of the coin toss to the situation of the 8 37/37 matches in a surname project. Each coin toss is totally unrelated to the previous x number of coin tosses and the next x number of coin tosses. That is not analagous to a situation where 8 men match exactly at 37 markers and have the same surname. If there is a possibility that the 8 men share just one paternal line, then you have to take into account that there is possibly only one "coin toss" among the 8 of them. And if they are indeed in the same paternal line, then each will very likely have the same 37 marker haplotype - one "coin toss" with the same result in each of the related men. In other words, the "coin toss" in this case is not 8 men, but the paternal line involved. If there were 8 different paternal lines involved in this comparison, then your analogy would hold true, otherwise not. That is the whole basis of DNA testing being applied to genealogy.

                          If the formula that FTDNA uses states that there is a 95% probability that the 8 have a MRCA in the last 7 generations, there is still a 5% chance that the MRCA is beyond 7 generations, let's say 10 generations. The formula is computed taking an average mutation rate observed in various studies and projects. As an average, there are many around the average mutation rate and as you move further from the average, there are much less at either the slow mutating and fast mutating end of the spectrum. What we are probably seeing in your paternal line is slower than normal mutation, so you have to look back more generations than in 95% of 37/37 matches.

                          That was a bit meandering. Does it make sense?

                          Mike

                          Very well put Mike.

                          Comment


                          • #14
                            Originally posted by MMaddi
                            Here's the error in your reasoning about making an analogy of the coin toss to the situation of the 8 37/37 matches in a surname project. Each coin toss is totally unrelated to the previous x number of coin tosses and the next x number of coin tosses. That is not analagous to a situation where 8 men match exactly at 37 markers and have the same surname.
                            Mike is, of course, right on.

                            If you could put all eight men an an accurate family tree, then you could answer the question about the odds that they all match each other 37/37.

                            As it is, we must guess but it isn't hard to do. If they all have a common ancestor born around 1700 (a distinct possibility, I'd say, given the genealogical data we have), then how many births are we looking at? I'd guess around 64 (eight in each line) at most.

                            If the mutation rate is around .0028 per marker per transmission event (i.e. birth), then the odds of a mutation in any given birth is about 10%.

                            What are the odds that we'd see 64 births without any mutations? About 1-in-a-thousand.

                            But we probably aren't looking at 64 transmission events. I'd guess it is more like 42, making the odds more like 1-in-a-hundred which is not at all extreme.

                            Plus, and this is crucial, we are selectively looking at the 37/37 matches. Many of the other members of the project probably descend from the same man but exhibit more mutations than expected. Since the MRCA calculations are Bayesian calculations, excluding the non-perfect matches is a critical error.

                            For example, imagine if there are 16 folks in a project with the same common ancestor born in 1700. Eight are a 37/37 match with each other, but eight others are a GD of 1 or 2 away from the modal in various different ways (which is actually what the Pitman project looks like to me). There would be nothing peculiar about that at all.

                            Comment


                            • #15
                              I can see that if it happened that two subjects were positioned close enough in a family tree to be tested true in one of the comparisons then that it would cause some interference with the other comparisons.

                              Specifically, if (a,b) = true and if (a,c) = true then the odds of (b,c) = true is increased; conversely if (a,b) = true and (a,c) = false then the odds of (b,c) = false is increased. so if this type of effect is what you were describing then, then I agree that this is not exactly analogous to flipping coins. HOWEVER it would seem to me that the fact that any two subjects compared false would not negatively affect the odds of any subsequent comparison being true. right?? If anything, I would think it would INCREASE the chance of a true test in one of the subsequent comparisons.

                              Don't get me wrong, I concede to you the point that you have successfully torpedoed the coin flipping analogy. How then would you estimated the probability that 8 individuals that match 37/37 and have at least 2 generations of documented paternal genealogies would not discover a MRCA?

                              Thanks again for the help MMaddi!

                              Bert

                              Comment

                              Working...
                              X