On 5/31/2016 2:18 PM, taf via wrote: > On Tuesday, May 31, 2016 at 9:53:17 AM UTC-7, joe...@gmail.com wrote: >> Perhaps, but both have been useful. Y-DNA testing has been useful to connect individuals genealogically back 10,11, maybe 15 generations back or more. >> >> autosomal DNA testing has been useful to me to prove/connect to individuals on a must more recent timescale (5 or 6 generations back with confidence). >> >> So, Y-DNA has a lot more depth on the chart, but autosomal has much more bredth in lieu of depth. >> >> Both useful, but in different ways. > I think Andrew was talking about a different type of autosomal analysis than you are. He was talking about SNP analysis used to determine ethnic proportions - basically useless for genealogy unless you don't know if your grandfather was a Finn or an Italian and you know everyone else is neither. What is more useful is the autosomal SNP clustering analysis that looks at conserved islands of contiguous DNA and can tell you someone is related to you within about a half-dozen generations. While it has limits, that can be useful data, though it proves no specific connection. The generation estimates given by the testing companies can be very misleading, and I sometimes wonder if they are making the results sound as good as they can possibly justify. I had my autosomal DNA tested by FamilyTreeDNA, and I have not been fortunate to find many matches close enough to be easily traceable, but when I got my results, I at least thought that the relationship ranges would be close enough that a significant percentage would fall within the range specified. My closest match, and the only one over 100 centimorgans total, was estimated to be in the 2nd-3rd cousin range, but I have no genealogical information on the testee to check the accuracy of that range. I have 26 matches that are estimated by FTDNA to be in the 2nd-4th cousin range. Of these, I have no genealogical information at all on five of them. Of the remaining 21, I have traced the paper-trail relationships for nine of them, which include one third cousin, one third cousin once removed, one double fourth cousin, four fourth cousins (one with very probable but unconfirmed additional matches further back), one fourth cousin once-removed, and one fifth cousin once removed. For two of the remaining "2nd-4th" matches, I have only very limited genealogical data, enough to rule out any second cousin or closer relation and no reason to expect a relationship as close as fourth cousin. For the remaining ten allegedly 2nd-4th cousin matches, I have enough information on their ancestry to rule out any relationship closer than fifth cousin with reasonable confidence (assuming acceptable accuracy in the information the testees reported), and no reason to suspect that any of them are even that closely related. (I have a paper-trail tenth cousin relationship with one of them, and a plausible conjecture for a sixth cousin relationship with another.) One or two of them might have a closer relationship due to a paper trail mistake or a "non-paternal event" but it is improbable that all of them do. Based on my confirmed matches so far, my own paper trail ancestry appears to be genetically accurate for at least 7 out of 8 great-grandparents (including all four great-grandfathers) and at least 11 out of 16 g-g-grandparents (including 5 g-g-grandfathers), making a "non-paternal event" on my end rather unlikely in this time frame. (Also, I saw no geographical or surname matches that looked promising in the appropriate time frame.) Thus, if the seven cases where I have inadequate information are left out, only seven out of nineteen of my supposed "2nd-4th" cousin matches lie within the estimated range, and not a single one lies within the lower part of that range. I never expected ALL of the matches to lie withing the estimated range, but I assumed that the ranges would at least be reported accurately enough that over 50% would fall within the reported range. When I read some of the more careful mathematical analyses available on the Internet, it was clear that given the total number of matching centimorgans in these cases, it was expected that a significant number of the relationships would be well beyond the fourth cousin range. Needless to say, I found this very annoying, as I did a significant amount of digging looking for paper trail matches which I would not have done if the probable relationship ranges had been more accurately reported. Are the testing companies deliberately using a simplistic mathematical model which makes the matches look closer than they really are? Stewart Baldwin