GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2003-05 > 1052339700


From: "John F. Chandler" <>
Subject: Re: [DNA] moving the spot
Date: Wed, 7 May 2003 15:35 EST
In-Reply-To: fauxdk@yahoo.com message <20030507034237.88342.qmail@web41212.mail.yahoo.com> of Tue, 6 May 2003 21:42:28 -0600


David wrote:
> Ok John and Grant, I think I am catching the drift of the discussion
> here. I know that fewer markers are available in the DNAPrint test to
> differentiate the EA and the NNA, and that they do admit that of all
> groups it is most difficult to distinguish between EA and NNA - but
> apparently that this is not a problem because................. Would
> your arguments / interpretations serve to explain how Ann and her sister
> obtained values of approximately 10% EA and NNA respectively?

Bear in mind that the experiment we were talking about is aimed at
understanding the effects of possible systematic errors in the model.
Although systematic errors could be a contributing factor in Ann's
case, they are not the main suspect, simply because systematic errors
would tend to affect Ann and her sister in roughly the same way. I
haven't looked at their triangle plots, but I assume the statistical
uncertainties are wide enough to explain most or all of the
discrepancy.

> looking at my mother - in - law's black and blue confidence interval
> lines around her 10% NNA, they hugged the MLE tightly, staying away from
> the EA line.

In another note, you remarked that these contours looked like semicircles
to you, implying that they intersect the side at right angles. If so,
that means the MLE as shown is absolutely the MLE. However, if they
are instead short arcs that intersect the side obliquely, then the
maximum likelihood is off scale. Another thing: I notice you say nothing
about the shape or position of the yellow ring, but that's the one you
need to concentrate on to understand the possibilities. Does it also
stay away from the EUR-EAS line?

> However both my uncle and his cousin's black and blue
> bands were virtually equidistant along the NNA and the EA lines, yet the
> former scored about 10% EA and the latter about 10% NNA. Is it just
> simply too difficult to call differences between EA and NNA at values in
> the 10% range, to the extent that slight differences lead the algori!
> thm to make a call in one direction or the other? David.

It's not so much the 10% range that makes it hard to call -- it's just
the particular genotypes that have alleles with ambiguous "meanings".
The size of the rings depends, in a complicated way, on the details of
which alleles show up and which ones don't.

John Chandler

This thread: