No announcement yet.

TIP Report

  • Filter
  • Time
  • Show
Clear All
new posts

  • TIP Report

    I'm having a problem understanding a report.

    I have two second cousins. One did the y67 and the other the y111. Their ancestry has been documented for over a hundred years. I am also a second cousin, female. We've all been documented through atDNA as 2nd cousins. There seems to not be any doubt. We, also, all match many, many others from the same surname atDNA group.

    Here's the question. Using both my male cousins results on the Y67 and using the Tip Report the indication is that their common male ancestor has a 70.48 % chance to be at four generations. Since we know they have the same great grandfather shouldn't their common ancestor be 100% or at least near that at three generations. Even at eight generations the estimate is at 94.78 % for a common ancestor. We know their common male ancestor is at three generations.

    Is their a problem with the Tip Report?

  • #2
    You must remember they are using probability based on 1,000s of people. You are basing your question based on two or three people. There is nothing wrong with their report.

    If you have 10 people in a room. 5 of them are 5 feet tall and 5 of them are 6 feet tall, the average height of the 10 people is 5.5 feet. Since none of the people are 5.5 feet tall are you going to say the average is wrong?


    • #3
      The TiP report assumes that you have NO OTHER EVIDENCE apart from the Y DNA results at the indicated level (Y 67 in this case). You can interpret this as a statement of how likely it is that two men, randomly chosen, who match to the indicated degree are related within a specified number of generations. The statement refers to the percentages you would get if you looked at a very large number of pairs of men. But those probabilities go out the window as soon as you know more, as you do. A better way to view the results, in light of your other paper trail as well as strong genetic evidence, is that there is nothing in the Y DNA evidence to cause you to doubt the autosomal DNA evidence and your paper trail. If, instead, the TiP report told you the odds were only 2 percent, then you would have to reexamine all of your other evidence.


      • #4
        Thanks for the replies. Now the report makes more sense. On a different note a project leader also pointed out to me that the tip report would give a higher percentage depending on a fast changing marker. For instance with another one of my kits the person has a GD of 3 (at Y111) with several men but the project leader said he was probably closer to one of those men because of a non match on a fast changing marker.


        • #5
          I have a related question. Does the TiP report factor in the timing of downstream mutations? I am trying to compare two viable genealogical hypotheses. One is supported by autosomal evidence, one is not. The first scenario also involves a known Y DNA match with a well documented genealogy. The problem is that this YDNA match is downstream of my haplogroup. I have seen one website indicate the mutations separating the related haplogroups could have happened as recently as 200 years ago (well after the genealogical wall I am trying to break through). The problem is that the TiP report only gives about a 50% chance at this level with my YDNA match. Interestingly, under this hypothesis, another match would share a common ancestor at the prior generation. TiP indicates 77% of a common ancestor at that level. (Those two gentlemen are in the same downstream haplogroup).

          I am beginning to think I am asking the TiP to do too heavy of a lift. For instance, the TiP report seems to indicate a common ancestor with another match by generation 8 or so (if I don't exclude any generations), when his genealogy indicates it may not happen until generation 15 or so.

          It seems that I am having to weigh the signficant autosomal evidence against the TiP predictions. I am leaning toward the autosomal evidence. Should I reconsider?


          • #6
            Originally posted by Doug McCullough View Post
            I have a related question. Does the TiP report factor in the timing of downstream mutations? . . .
            Sorry, for interrupting, but I don't think the TiP report for STR matches factors in the results for any SNP tests. I've been able to closely approximate TiP results solely by reference to STR profiles, even though some of the kits I administer have varying levels of SNP testing (highest being BY700).

            I'm not sure I'm 100% following every nuance of your description of the situation, so the following comment may not necessarily be of interest to you. But the 'sweet spot' for TMRCA calculation curves is at the 50% mark--the single most likely TMRCA, although the level of randomness is formidable. 77% confidence at date X, where X is the hypothesized birth year of your MRCA, is not necessarily 'better' than 50% confidence at that same date. In fact, technically, 77% is usually a little bit worse, when you're interested in precision in an absolute sense. 77% indicates that more than the expected number of mutations for a MRCA born around X have been observed, whereas 50% indicates that exactly the expected number of mutations have been observed. 77% isn't necessarily 'better' than a 23% confidence score, either, because they're equally distant from the 50% sweet spot.

            But if you define your parameters as a one-sided concern that MRCA was born no more recently than year X, then, yeah, I guess 77% would be better. You'd know that only 23% (i.e., 100%-77%=23%) of all points on this probability curve are more recent than year X. I guess it's all very specific to the full context of the hypothesis that you're working on.

            But frankly, 50%, 77%, 23%, they're all in play. They're respectable scores that don't really suggest that the predictions are in serious conflict. Really, if there is a standard rule of thumb that researchers apply, it seems to be that any data falling outside the 95% two-sided confidence interval (i.e., 100%-95%=5%; 5%/2= or lower than the 2.5% lower boundary; 100%-2.5%=or higher than the 97.5% upper boundary) is suspect or at least considered an outlier unrepresentative of the likely range of true relationships.

            TMRCA calculations follow what they call a binomial distribution, which is almost identical to the fully symmetrical normal distribution, with just a somewhat longer right tail (i.e., more possible really old TMRCAs on the continuum than really recent TMRCAs). So probably just under 68% of all possible TMRCAs fall within 1 standard deviation of the sweet spot (i.e., 50%). So individual matches showing up between the 16% to the 84% confidence levels are probably not a big whoop (i.e., 68%/2=34%; 50%-34%=16% lower bound; 50%+34%=84% upper bound).

            That said, it sounds like a more recent founding date for this downstream clade is more favorable to your preferred hypothesis. Dating of SNPs still seems to be in the wild-west pre-consensus stage, IMO. So it's probably not something to get hung up about. But based on an admittedly small number of case studies, I am inclined to believe that all the popularly discussed mutation rates that I've seen are much slower than is actually the case. At the moment--subject to change upon clarification/further evidence, of course--I'd say the true population average rate is somewhere around 1 SNP in 86 years at the legacy BY500 resolution or 1 SNP at 57.333 years at the BY700 resolution. So it sounds like my ideas might be consistent with your preferred theory.

            If I understand where you're coming from, which I may not.

            But just one final thought, because the time horizon with more recent relationships is so more shorter than the horizon for remote relationships, the typical difference between actual observations along the estimated probability curve tends to appear more jarring to our natural way of thinking of numbers in a fixed sequence of 1's, (i.e.,1, 2, 3, 4 , etc. and so forth). When considered from the point of view of a logarithmic sequence (i.e., 10, 100, 1000, 10000, etc., etc.) they're not. Just another reason not to sweat apparent diversity in reported confidence levels for most matches.
            Last edited by benowicz; 27 June 2021, 03:40 PM.


            • #7
              Thank you Benowicz! This is very helpful, even if this old liberal arts major doesn't fully understand it at just one reading!