GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2008-09 > 1222288738


From: "David Faux" <>
Subject: Re: [DNA] Variance - was What shall R1b1c call themselves now?
Date: Wed, 24 Sep 2008 13:38:58 -0700
References: <200809241236.m8OCaoWR016154@mail.rootsweb.com><REME20080924162212@alum.mit.edu>
In-Reply-To: <REME20080924162212@alum.mit.edu>


Yes, and this is precisely what is wrong (well one of the things) with the
nouveau dating for M269 and subclades. People are using my database (or
that of Charles Kerchner) to estimate a date for R-U152. I know my database
better than anyone. It is hopelessly biased and flawed for the purposes of
calculating node or SNP dates. There is a huge skewing toward the British
Isles which is distorting the data in ways that may be giving an
artificially young date for R-U152. Now if one was only interested in the
British Isles, then it would appear to offer a fair opportunity to estimate
the TMRCA for that group, but only that group within R-U152.

Second, we now find that R-U152 has ancestral and derived clades (in
essence, a new haplogroup has emerged in L2). Thus combining L2 ancestral
and L2 derived can be justified depending on ultimate goals. However it
also confounds the data and we are hopeless left adrift in our quest to find
the "age" of R-U152 since we are left with unknown number of surviving
lineages. Separate analyses will need to be done of Continental L2+ and -.
Unfortunately to date, the ancestral variety appears to be relatively small
in number and so sample size is going to factor into our uncertainty.

David K. Faux.


On 9/24/08, John Chandler <> wrote:
>
> 4. However, it matters how the data are obtained. If you assemble a
> large collection of data with a systematic bias of some kind, then
> the calculated estimate can be seriously wrong, way beyond the
> limits of the calculated uncertainty. Note that the term "bias"
> has a technical meaning here -- it is simply a condition where the
> data turn out not to be evenly representative of the whole
> population. For example, suppose you did a study of nutrition
> intending to compare college students with high-school dropouts
> by comparing the heights of 20-year-olds. You would get a very
> distorted picture if the college sample were drawn from the
> basketball teams.
>
> John Chandler
>
>

This thread: