Archiver > GENEALOGY-DNA > 2004-01 > 1073060582

From: "Debby Peare" <>
Subject: Re: [DNA] By George, I think I've got it! LOL
Date: Fri, 2 Jan 2004 11:23:02 -0500 (Eastern Standard Time)
References: <>


Let's just say you're beginning to get it.
No, I do get it. I just didn't elaboate further.

The yellow confidence contour on your triangle chart is NOT a band of
"least likely" possible values, though it does represent a "scientific
level of confidence." It's a line made up of points each of which is 10
times less likely to be your true value than your MLE. A confidence
interval of 100 times less likely, or 1,000 times less likely, or
whatever, could be added to the triangle chart, and any point on those
contours could possibly be your true value, rather than the MLE. Very
unlikely, but still possible.
True. Poor choice of words. But once sent! LOL
What I meant to say was the least one shown. I know that there could be more
beyond the 10, any within the contours and with the contours, you have the
range of - less likely as well.

Furthermore, it is just as likely that your true minor ancestry
percentage lies "somewhere" among any ten points on your yellow
confidence contour as at the point of your MLE. And, it is MORE likely
that your true value lies "somewhere" among any eleven or more points on
the yellow line than at your MLE. It is twice as likely that your true
value lies "somewhere" among any twenty points on the yellow contour
than at your MLE. So, in most cases, it is more likely that your true
minor ancestry lies "somewhere" among the possible values on the yellow
confidence interval which lie BELOW 0%, than at your MLE!
But THERE is the rub!! You take the 2, 5 or 10 times "less likely" (or any
point thereon or between) and say, it's MORE likely that my "true" MLE is
anywhere but where it is. What is truer said is the MLE is the BEST estimate
of what the porportions are with a small chance that they are something else
(thus the contours) because of the testing of only the small amount of the
variable genome and because the MLE is a statistical estimate, thus the MLE
COULD move a small bit. It's not twice as likely that my true value lies in
the contours, it is 2 fold less likely than the MLE on the first contour, 5
fold less likely on the second and 10 fold less likely on the third, etc.
The further you go from the MLE, the further it is from the MAXIMUM ESTIMATE
of your true value. In fact, the best estimate on the data on repeated
measurements (they use 5 samples on each run as internal controls) tested
has a 3-4% variation for MOST determinations with individuals that have
failed markers. So, in other words, if any minor ancestry is 4% or less, it
d be debateable if that minority is significant.

The 5% error is supposedly the maximum lab error. There is an as yet
undetermined systematic error rate (due to faulty sampling), which MUST
be quite high, since someone with no NA ancestry can be given a 10% NA
MLE and a gray (2 times less likely) confidence interval which never
drops to 0%.
And here again, I say, if there's smoke, there most likely is fire and if I
had that 10% NA, I'd be asking many questions. Just like I am my 12%S-SA.
You saw my triangle and know that the contours indent towards the S-SA
considerably lending even more credibility to the possibility being more
likely than not that the 12% S-SA is a good bet being true. What was your
mother's contours? And I say this, it is quite possible that they were wrong
in your mother's case, but also quite possible with it being 10% that they
are correct and someone in your family just didn't know or didn't tell, IMHO

But then again, I too am no scientist, just a very intelligent person who
once has grabbed the info can intelligently debate the issues.

I'm not a scientist, Debby, but I can see some of what's wrong with this
test. Pessimistic? You bet!

Ray Whritenour
Scientificly, it's good to be pessimistic. But even then, you must present
the facts correctly, and your view of the MLE most likely being in the
scientific level of confidence (10 fold less likely than the MLE) is to me,
being TOO pessimistic and I submit that the truer value is most likely
inbetween the MLE and the 2 fold less likely than the MLE. I have more faith
in the newer testing and it's algorithm than you. And I suspect that future
tests will bear my assumptions correct.

Warm Regards from Maine,
Debby Peare

This thread: