GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2004-01 > 1075066460


From:
Subject: Re: [DNA] PA Deutsch Project -- more DNAPrint Test Results uploadedto Log
Date: Sun, 25 Jan 2004 16:34:20 -0500 (EST)
References: <4012ED01.8010907@kerchner.com> <4012F4F7.3070107@kerchner.com>
In-Reply-To: <4012F4F7.3070107@kerchner.com> (message from Charles on Sat, 24Jan 2004 17:43:03 -0500)


Charles wrote:
> I am weak in statistics and my eyes rapidly glaze over after a few
> paragraphs. Would someone please comment on the statistical analysis
> procedures, process, and results of the analysis of my Log data, and the
> conclusions reached which were recently posted in the Accuracy and
> Precision section of the new AncestrybyDNA website.

A few comments.

It appears that the analysis focused selectively on aspects favorable
to DNAprint marketing. It should therefore be viewed as advertising,
rather than as scientific analysis.

Nonetheless, the results are probably correct, as far as they go. I
haven't bothered to check the numbers. Assuming the numbers are
correct, the table comparing averages of expected vs measured
percentages is the closest thing yet to a study of possible systematic
error in the DNAprint procedure, along the lines I proposed last year.
The limitation of this particular study is obvious: it is relatively
small, despite Charles' best efforts to encourage people to sign into
the DNAprint log. Being small, it is vulnerable to all kinds of
sampling biases, including the effect that Charles commented on
yesterday (that many of the comments seemed to reflect "expectations"
based on discussion of other DNAprint results, instead of their own
genealogical resources).

John Chandler

This thread: