GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2010-03 > 1267644403


From: (John Chandler)
Subject: Re: [DNA] DYS464X and DYF399X
Date: Wed, 3 Mar 2010 14:26:43 -0500
References: <992793705.1871541267582956506.JavaMail.root@sz0002a.westchester.pa.mail.comcast.net>,<67710699.1879681267584274324.JavaMail.root@sz0002a.westchester.pa.mail.comcast.net>,<fef9eee41003021926h2961ba9fl4b32425b7b40081f@mail.gmail.com><SNT115-W325A6880C2D5D5479D9156CC3A0@phx.gbl>
In-Reply-To: <SNT115-W325A6880C2D5D5479D9156CC3A0@phx.gbl> (message fromSteven Bird on Wed, 3 Mar 2010 10:35:52 -0500)


Steve wrote:
> Eh? I don't follow. At 67 markers, a close match is a close match.
> Why would it matter how many have matched in total? If you were to
> match 64 out of 67 all with one step differences, for example, it
> would indicate a common ancestor within the past few hundred years,
> even if there were fifty men in this cluster. They would all be
> related within a genealogical time frame.

You need to beware of sliding into the Frudakis definition of "a few".
For a 64/67 match with three one-step offsets, the 95% confidence interval
of TMRCA goes back six centuries in the absence of any correction for
skewed priors. If there were 50 men in the same cluster with you, and
if the modal haplotype of the cluster "split the difference" between
you and the other guy, then prior correction would just about double that
interval to twelve centuries.

The point is that your common ancestor almost surely had the modal
haplotype under the conditions I described, and the proportion of
men in the cluster who actually have the modal haplotype is decaying
exponentially with time. In other words, the whole probability
distribution gets pushed back in time because the prior probability
strongly favors an earlier MRCA over a later one.

On the other hand, if you are 67/67, then you gain two ways. First,
the uncorrected 95% confidence interval shrinks to two and a half
centuries. Second, the likelihood grows that you (and your MRCA) do
*not* have the modal haplotype of the cluster, and, hence, the prior
correction probably goes away.

John Chandler


This thread: