GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2003-09 > 1062473635


From: David Faux <>
Subject: Re: [DNA] Cecelia's Mother's DNAPrint 2.0 Results - a big surprise
Date: Mon, 1 Sep 2003 20:34:03 -0700 (PDT)
In-Reply-To: <000d01c370ec$b3fc8750$a99d72d1@Garnet>


Cecelia:

If you are correct - and curiously I made a statement about their algorithm being potentially terribly flawed in a post sent about the same time as yours - then this is truly disturbing. I am running a whole series of tests on descendants of each of the 4 children of my documented Native American ancestor. Should you perchance be correct (and I fear you may be) then I will be comparing apples to oranges, and my results (which I planned to include in a 3500 name book on the Young family) are potentially worthless and perhaps meaningless.

Considering the number of individuals on this List who have reported minority Native American, but have not a shred of documentation to support the finding (plus a genealogy that made the findings very unlikely) I am beginning to wonder whether the algorithm was "amended" at some point to give the customer what they want - a little bit of Native American. Now the algorithm would not work 100% of the time to effect this purpose, so there would be some individuals who might even come up 100% Indo European - these results basically being a false negative - however playing the numbers game it would give they and their customer's what they want more times than not.

Another hypothesis is that they were well aware that they were getting Native American signals where they should not be and have altered the algorithm to address this matter. Tweaking the algorithm is fine as long as the customer is informed of the change and the rationale behind this move. In addition it should be done in conjunction with the data from people such as myself with a fully documented genealogy. Here, in testing 10 members of my family (and others in the same boat), if the algorithm is on the money, then the results should follow the laws of genetics. Otherwise this is a very shabby move and I think that it is time I brought my project to a halt until the DNAPrint people can explain what the heck they are doing behind the scenes.

There has often been a veil of secrecy involving this test which has led to much speculation and few hard facts. When time permits I will thrash out the matter with Tony and Matt at Ancestry by DNA but at the moment am too preoccupied with other matters. Perhaps someone would like to grill them a bit on whether there have been changes to the algorithm (and how many times this has occurred); and what we can expect from the 2.5 / 3.0 version (very little more than we have I fear since I have not heard that they have been doing extensive testing of representative samples from around the world - hope I am misinformed).

David F.

Cecelia Clancy <> wrote:
Hello All,

I got back my mother's DNAPrint 2.0 results. Her result was 100%
Indo-European.

This is a strange result, given that mine were (first test) 94% IE and 6%
Native American and later, (second test) 92% IE, 2% NA, and 6% East Asian.
Since my father immigrated straight from western Ireland (strongly
suggesting his results would have been 100% IE), and also because my mother
thinks that her father was "part Indian", I presumed that my NA and/or EA
signals were coming from my mother's side instead of my father's.

But now, DNAPrint estimates her to be 100% IE.

This can mean one of at least three things (or some combination):

1) DNAPrint is inaccurate to begin with when it comes to minority ancestry.
(Many people on this list now believe this.)

2) DNAPrint has made changes to the algorithm (formula) between my tests and
my mother's test. In other words, if my mother's data were run through the
algorithm used on my first and second tests - my mom would show up with
significant NA.

3) I have very unexpected NA and/or EA signals coming from my father's side.

But with this new result, it would be interesting to know what my father's
DNAPrint results would be, but that is impossible, for he is deceased. (Yes,
it might be still possible to get something with "his DNA" on it, but for
now, I'm presuming we won't be able to.)

But by comparing mine and my mother's results, I can partially reconstruct
my father's. And when I get more money, I'll get more of my sibblings
tested - with their results, I will be able to reconstruct more of my
father's. When I reconstruct much of it (I have seven sibblings - all of the
same father and mother), I'll ask DNAPrint to run the reconstructed results
through their formula - and use the same formula*S* they used to calculate
my first and second results.

Why do I write "formula*S*" above?

There is a very good reason - something I had forgotten about when I posted
awhile back about the differences between my first and second DNAPrint
results.

Here it is. I recall speaking on the phone with somebody at DNAPrint - to
inquire as to why after months, I have not yet had back the results of my
second test. They looked it up, and told me over the phone that this time,
there were no failures. (On my first test, there were several failures.) But
the rest of what that man said was very important - that I'm sharing here
for the first time.

He said that DNAPrint also changed the algorithm!

I forget his exact words, but they contained terms on the order of
"adjusted" or "changed" or "played around with" and "formula." (He did not
use the term "algorithm".)

So, the differences between my first and secon DNAPrint test are accounted
for not only by differences in the data itself, but also in the statistical
formula or algorithm they used to estimate my results!

I wonder now, if DNAPrint had again changed the algorithm sometime after my
second test and before my mother's test???? If so, would running my mother's
data through the same two formulas used for mine yield significant NA/EA
results for my mother? And would running my results through whatever
algorithm was used for my mother's results yield a 100% IE result for me???

It is possible that DNAPrint - after seeing in this forum (and elsewhere, no
doubt) the many complaints about NA and EA showing up unexpectedly - had
concluded privately that "Yeah, they're right. We gotta fix the algorithm."
And when and if they "fixed" it, they "fixed" it so the results would show
less NA and EA for people who have majority IE results?????

Regards,
Cecelia Clancy

Researching:

Blechinger,
Brabeck (with an umlaut),
Burnett,
Clancy,
Conneely or Connelly,
Conway,
Dunlay
Jacob,
Kearney,
King,
Loritzen (and spelling variations),
Mader,
Maher,
Meder,
Meeder,
Miller,
Mueller (with or without an umlaut),
O'Toole,
Plechinger,
Prabeck,
Schmotzer,
Schuster,
Shumate,
Toole, and some more I have not listed yet.



==============================
To join Ancestry.com and access our 1.2 billion online genealogy records, go to:
http://www.ancestry.com/rd/redir.asp?targetid=571&sourceid=1237


Dr. David K. Faux, P.O. Box 192, Seal Beach, CA, 90740, USA



www.davidkfaux.org


This thread: