Archiver > GENEALOGY-DNA > 2006-08 > 1156297684

From: (John Chandler)
Subject: Re: [DNA] When is 25/26 not enough?..
Date: Tue, 22 Aug 2006 21:48:04 -0400 (EDT)
References: <>
In-Reply-To: <>(

John wrote:
> MY REPLY: I don't follow this. If I have one dice, I understand there is 1
> chance in 6 of throwing a six. If I throw a second dice, I understand there
> is 1 chance in six of throwing a six. If I throw two dice together, I
> understood there was 1 chance in 36 (6x6) of throwing 2 sixes.

You are assuming, with justification, that the two dice are independent.
You cannot make that assumption about two markers in one haplotype.
It's that simple. This is why the forensic crowd gets very excited
and distressed if someone challenges the assumption of equilibrium in
the autosomal markers. The whole basis of using CODIS markers to
identify criminal suspects or prove/disprove paternity is the assumption
that the probability distributions of the various markers are all
independent, so that they can multiply the frequencies together and
come up with an astronomical number for the likelihood that any two
people would have the same values by accident.

This is also why Y DNA markers are *not* used for identification.

> MY REPLY: I didn't single out those markers because I found I matched
> another person on both those markers. I singled out the markers for
> specific attention, because I found "my" results were rare on those markers.

You picked on those markers only because the values were rare *and*
you matched, and you would just as quickly have picked on any other
markers that turned out in your case to have rare values and to match.
What is the next higher allele frequency among your 67?

By the way, if these are indeed slow-moving markers, then they could
easily be clade markers for clades that happen to be very poorly
sampled at present among the very limited pool of Y67 testees. In
short, their rarity could easily be dependent on geography.

> Then I said, if I test "on those particular markers", another person of my
> surname, from a particular family which I suspected of having a deep
> ancestry connection, from a nearby area to my ancestral area, than randomly,
> there should be 1/20,000 chance of them matching both markers, "if we are
> not related".

As I've pointed out before, the correct number is not 1/20,000, but
1/200, even if you assume the apparent rarity is real. What you need
to do is test some others who also come from the target area and see
if *they* have an elevated tendency to share these "rare" values.
Perhaps you have discovered a new clade, but the disparities on the
other markers should be telling you that it's not a young clade.

John Chandler

This thread: