GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2003-07 > 1057090320


From: "John F. Chandler" <>
Subject: Re: [DNA] Ancestry test - a must-read article URL
Date: Tue, 1 Jul 2003 16:12 EDT
In-Reply-To: Circles54@webtv.net message <5531-3F010852-1617@storefull-2173.public.lawson.webtv.net> of Mon, 30 Jun 2003 22:04:21 -0600


Ray wrote:
> Look at anybody's triangle plot, who has an MLE showing minor NA
> ancestry, and at least two of the confidence contours run from 0% NA to
> as much as 40% NA.

It's not clear what you're saying here: are these two conditions, or are
they a condition and a conclusion? It sounds more like the latter, but
it's not so. For anybody with at least 30% NAM, none of the rings is
likely to reach 0%.

> Most have many points which plot below 0%. And, as
> DNAPrint tells us, any one of these points is your possible true value.

Strictly speaking, the rings have points which WOULD plot below 0%,
except that the plots are always truncated at the edges of the triangle.
It would really be a big help if the plots were extended all the way
around, but I suppose the software is not set up to do that.

> In most cases, there are so many points which plot "below"
> 0% that one of those points is MORE LIKELY to be your true percentage
> than your MLE!

That's a little twisted. When you say "one of those points", it no
longer is relevant how many of them there are. Each point has its own
likelihood of being the truth, and the likelihood for any small area
(say the size of the red dot) is inherently small. If you're making a
comparison between the small red dot and the extensive swath that WOULD
be plotted in negative territory, then you have to specify the whole
swath instead of letting "one point" stand for the whole. On the other
hand, you COULD be making a completely different point, which is that
the red dot is not the absolute maximum likelihood in cases where the
real maximum is off-scale. In such cases, the red dot merely marks
the highest likelihood found within the triangle, and the true MLE (off
the edge in negative territory) is indeed more "likely" than the red
dot. The problem is, of course, that the percentages are a
mathematical abstraction, rather than a physical measurement of "blood".
As such, there is no real boundary at the edge of the triangle.

John Chandler

This thread: