Bernie, Although the reconstructed STR tree is not the same as the genealogical tree, it is not that far off in simulations I have tried. Early on, I created some synthetic genealogies over a few hundred generations with a Poisson distribution for the number of sons born to each father (so that various lines would go extinct) and with various permutations of population growth models that I thought were realistic. >From those synthetic genealogies, I then simulation how 67-markers, with assumed mutation rates would propagate through to the present. Then I tried the hierarchical clustering algorithm to reconstruct the STR tree from that simulated data. I came to the conclusion that I could acceptably approximate the synthetic genealogy albeit with some annoying convergence in some places. But visually the STR tree often looked remarkably similar to the synthetic genealogical tree after one removed the extinct lines and if one looked at the lines were convergence wasn't an issue (which it often wasn't too bad after a few hundred generations). Probably should write that up one day. (I am doing all of this work in my spare time as a hobby.) So the STR Tree for I1 and I2 that I give in the PDF file, I have a feeling is not an unreasonable representation of the actual genealogy. Can't prove that in the actual case at hand, but the simulations do suggest that in principle with the noted caveats. (After more than many hundreds of generations, and with the mutation rates I assumed, things may start to get worse, but I couldn't simulate that due to computer runtime constraints). Also note that SNP alleles should exactly follow the actual genealogy, because they are usually unique events, and in the reconstructed STR tree I did for I1 and I2, the SNP locations are mostly (but not always) grouped in non-overlapping areas. Terry On Sun, Mar 4, 2012 at 1:15 AM, Bernie Cullen <berniecullen@gmail.com>wrote: > Terry, > > Your clustering method may work well, but as you say some branches may > converge. So what's the advantage of presenting your tree when we know > its basic structure doesn't represent history? What useful information > do your different branches give us? Why not keep your tree and branch > codes in the background (let someone click to it if he wants) and just > present the results (matches in someone's cluster, maps, recommended > SNPs)? > Or am I missing something useful about the tree? > > Bernie > > Terry tdrobb@gmail.com wrote: > Finally, don't expect total accuracy. The STR 67-marker Tree that is > constructed, can only ever be an approximation to the true genealogical > tree. Mistakes necessarily happen due to chance convergence of STR results. > All we get to see are the leaves (people) at the end of the branches, and > if we are lucky a bunch ("cluster") of leaves will all come from the same > branch. But sometimes two distinct branches will have their leaves > intertwined ("convergence"), which is the main complication to be aware of > when reading the results. Some spurious results are to be expected. > > ------------------------------- > To unsubscribe from the list, please send an email to > Y-DNA-HAPLOGROUP-I-request@rootsweb.com with the word 'unsubscribe' > without the quotes in the subject and the body of the message >