GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2005-03 > 1110349365


From: (John Chandler)
Subject: Re: [DNA] Re ; DNA Print test
Date: Wed, 9 Mar 2005 01:22:45 -0500 (EST)
References: <5862-422E6132-892@storefull-3134.bay.webtv.net>
In-Reply-To: <5862-422E6132-892@storefull-3134.bay.webtv.net>(Circles54@webtv.net)


Ray wrote:
> Without even looking at the triangle chart, a 9% "EA" test result means
> that the average "European" has about a 91% probability of having "SOME
> LEVEL" of "EA" ancestry (according to DNAPrint genomics, Inc.).

I'm afraid you've just confused BJ more. The 9% EAS result is just
that: the best guess of the admixture, based on the reference
samples collected by DNAprint, is 9% East Asian. The 91% EUR result
means the best guess is 91% European.

It does NOT mean 91% likelihood of SOME level of Asian. By that
chain of reasoning, a test report of 20% EAS + 80% EUR would have
a smaller likelihood of any real Asian component than the 9&91 report!

> In a "European" individual, this most likely shows deep "EA" ancestry,
> rather than recent "EA" ancestry; perhaps stemming from Hun or Mongol
> invasions of Europe.

Huns and Mongols aren't all that deep. You have to consider ancestry
all the way back to the original separation of populations coming out
of Africa.

> Of course, this also means that there is about a
> 9% chance that such an average "European" has "ZERO" percent "EA"
> ancestry.

Aside from the mistaken logic (it's not a 9% chance), this is a very
important point. The typical uncertainty on the Asian estimate for
DNAprint 2.0 is somewhere in the range of 12-15 percentage points, as
you can see from the triangle plots. That means a person with a report
of 9% Asian has about one chance in four of having zero Asian ancestry.
9% is still the best guess, but it could be wrong nonetheless.

> On your triangle chart, view it so the "EA" corner forms the apex of the
> triangle. The red dot (MLE) should show up exactly 9% of the distance
> from the base of the triangle to the apex. Around this red dot are
> three circles, called "confidence intervals." The first such circle is
> black. There is about an 88% probability that your REAL or TRUE "EA"
> ancestry falls SOMEWHERE within this black ring.

No. The black ring is the 50% confidence contour. (More precisely, it
is the max/2 probability density contour, but it just so happens that
these are the same thing for a two-dimensional normal distribution.)
That means the probability is about 1/2 that the real admixture lies
somewhere inside the black ring. And the probability is about 1/2
that the real answer lies outside the black ring.

John Chandler


This thread: