In a message dated Wed, 2 Jun 2010 20:31:31 EDT, [email protected] writes: >> I am having a hard time "getting" this mention of "1/2 cousin". Maybe someone can explain it to me. << Not complicated at all. Let's take my case as an example: My father had a half-brother. That man's daughter and I aren't "full" first cousins. We are "one-half" first cousins. Then that daughter's children and my children are 1/2 second cousins. And the daughter's grandchildren and my grandchildren are 1/2 third cousins. >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> In a message dated Tue, 1 Jun 2010 11:58:29 EDT, [email protected] writes: >> I don't know too much about overall percentages -- at least not with the algorithm 23andMe uses. For example, I have a proven 2nd cousin once removed and our overall match is 1.76% and her son (my 3rd cousin) matches 1.02% total DNA according to 23andMe << The "theoretical" percentages are very straightforward: Everybody shares "on average" a theoretical one-half of DNA with a parent or sibling. Then with a full first cousin, the average theoretical share is one-fourth of 50%, or 12.5%. Go back another generation, and we divide by four again: The theoretical average share with a second cousin is 12.5/4 = 3.125%. And the theoretical share with a third cousin is 0.78%. For a 1/2 cousin of any degree, just divide the theoretical share by two. Ditto for a "once removed" cousin. So the theoretical share with a 2nd cousin once removed is 1.56%. Now in the real world, the "actual" shares will vary statistically around the theoretical averages, and the variation becomes a greater and greater problem as we go back to earlier and earlier generations. This increasing problem is due not only to the random nature of DNA recombination, but it's also due to the fact that most of our first, second, third and fourth cousins also are probably seventh, eighth, ninth and/or tenth cousins via entirely different ancestral lines -- lines that probably will never be identified but that nonetheless may have contributed "extra" DNA to the matches we observe with close cousins. As to any "algorithm" that 23andme might use, I suppose the reference is to confidence intervals for determining whether an observed percentage falls within the statistically probable range for a certain relationship. My guess is that 23andme probably applies something like a "ninety-five per cent" or "two sigma" rule, based on analyses of matching percentages for representative samples of known cousins. But whatever confidence interval is applied, Joan's percentage match of 1.76 with a proven 2nd cousin once removed must certainly be "close enough" to the theoretical percentage 1.56. Best regards, Jim Brown (James Armistead Brown, Jr.) Administrator, Brown DNA Study Administrator, Knox DNA Project