Ann, Thanks for sharing your experience. Have you looked at the raw data on those long segments (over 10 cM) to determine if they really are IBS? I find those lengths very surprising. With your data, you noted that about 95 percent of the time what I said earlier will hold true, but that other 5 percent is troubling. I certainly would like us all to be able to be confident of the validity of the 10 cM blocks and above, if not the 7 cM and above. CeCe Sent via BlackBerry by AT&T -----Original Message----- From: Ann Turner <[email protected]> Date: Thu, 17 Nov 2011 12:19:52 To: <[email protected]> Subject: Re: [AUTOSOMAL-DNA] Fundamental autosomal DNA question One way to study the problem is with father/mother/child trio data. If the child has a match not found in either parent, it is possible (indeed I'd say likely) that the segment is Identical by State, not Identical by Descent. Ancestry Finder at 23andMe lets you look at segments down to 5 cM / 700 SNPs. For actual numbers in one case, my son has a total of 231 segments listed with names attached. 67 of those are found only in the child (29%). The breakdown by segment size is 4/76 > 7 cM, longest 11.1 cM (5%) [This seems consistent with a 95% confidence interval] 7/37 6-7 cM (19%) 56/119 < 6 cM (47%) FTDNA seems to have a cutoff of 7.7 cM / 500 SNPs (that's by empirical observation -- has anyone found a shorter longest segment?). I have data for one father/mother/child trio that seems particularly dicey, with 22/100 matches in the child not found in either parent. The segment sizes ranged from 7.7 to 12.90, with a median of 8. Other trios I've looked at seem to run about 15%. Ann Turner On Wed, Nov 16, 2011 at 1:02 PM, CeCe Moore <[email protected]> wrote: > > When I say "legitimately" match, I mean an IBD segment that would meet the > company's matching threshold, not a segment made up of stretches of IBD and > IBS pieced together. > > I'm not saying that it is impossible for a match over ~7 cMs to be IBS, > but very unlikely. (I was going to write ~10 cMs in my response, but > thought that might be a bit conservative.) I know that it is theoretically > possible, but in my thousands of hours of atDNA research, I have yet to see > proof of one. Can you please show me an example of an IBS match over 7 cMs? > (I know that Ann Turner has seen lots of them between 5cMs-7cMs. We'll see > if she responds with a larger one.) > > If enough of a portion of the segment is IBS that it doesn't show up in > your parent (i.e.- make the company's threshold), then it will obviously > reduce the authentic match to a point that is not worth pursuing. I have > chased far too many matches under 7 cMs to recommend doing the same. > ------------------------------- To unsubscribe from the list, please send an email to [email protected] with the word 'unsubscribe' without the quotes in the subject and the body of the message

No, unfortunately. I need raw data with father/mother/child trios on both sides of the match so I can use phased data. That's not easy to come by. Ann On Thu, Nov 17, 2011 at 7:41 AM, CeCe Moore <[email protected]> wrote: > Ann, > Thanks for sharing your experience. > Have you looked at the raw data on those long segments (over 10 cM) to > determine if they really are IBS? I find those lengths very surprising. > With your data, you noted that about 95 percent of the time what I said > earlier will hold true, but that other 5 percent is troubling. I certainly > would like us all to be able to be confident of the validity of the 10 cM > blocks and above, if not the 7 cM and above. > CeCe > Sent via BlackBerry by AT&T > > -----Original Message----- > From: Ann Turner <[email protected]> > Date: Thu, 17 Nov 2011 12:19:52 > To: <[email protected]> > Subject: Re: [AUTOSOMAL-DNA] Fundamental autosomal DNA question > > One way to study the problem is with father/mother/child trio data. If the > child has a match not found in either parent, it is possible (indeed I'd > say likely) that the segment is Identical by State, not Identical by > Descent. > > Ancestry Finder at 23andMe lets you look at segments down to 5 cM / 700 > SNPs. For actual numbers in one case, my son has a total of 231 segments > listed with names attached. 67 of those are found only in the child (29%). > The breakdown by segment size is > > 4/76 > 7 cM, longest 11.1 cM (5%) [This seems consistent with a 95% > confidence interval] > 7/37 6-7 cM (19%) > 56/119 < 6 cM (47%) > > FTDNA seems to have a cutoff of 7.7 cM / 500 SNPs (that's by empirical > observation -- has anyone found a shorter longest segment?). I have data > for one father/mother/child trio that seems particularly dicey, with 22/100 > matches in the child not found in either parent. The segment sizes ranged > from 7.7 to 12.90, with a median of 8. Other trios I've looked at seem to > run about 15%. > > Ann Turner > > On Wed, Nov 16, 2011 at 1:02 PM, CeCe Moore <[email protected]> wrote: > > > > > When I say "legitimately" match, I mean an IBD segment that would meet > the > > company's matching threshold, not a segment made up of stretches of IBD > and > > IBS pieced together. > > > > I'm not saying that it is impossible for a match over ~7 cMs to be IBS, > > but very unlikely. (I was going to write ~10 cMs in my response, but > > thought that might be a bit conservative.) I know that it is > theoretically > > possible, but in my thousands of hours of atDNA research, I have yet to > see > > proof of one. Can you please show me an example of an IBS match over 7 > cMs? > > (I know that Ann Turner has seen lots of them between 5cMs-7cMs. We'll > see > > if she responds with a larger one.) > > > > If enough of a portion of the segment is IBS that it doesn't show up in > > your parent (i.e.- make the company's threshold), then it will obviously > > reduce the authentic match to a point that is not worth pursuing. I have > > chased far too many matches under 7 cMs to recommend doing the same. > > > > ------------------------------- > To unsubscribe from the list, please send an email to > [email protected] with the word 'unsubscribe' without > the quotes in the subject and the body of the message > > > ------------------------------- > To unsubscribe from the list, please send an email to > [email protected] with the word 'unsubscribe' without > the quotes in the subject and the body of the message >