GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2006-07 > 1153687870


From: "M True" <>
Subject: Re: [DNA] Re: New median network diagram of R1b
Date: Sun, 23 Jul 2006 21:52:18 +0100
References: <20060723190903.90740.qmail@web34213.mail.mud.yahoo.com>


Hi Mike,

As the person who set up the diagram I think I should try to answer this one
as best as I can. I think the problem lies in the fact that the diagram
tries to "stretch" the 37 marker data for each individual to the limit in
terms of trying to find connections. There are some lines which seem to
come directly from the mode, presumably where the Fluxus program couldn't
make any connections whatsoever between individuals.

To create the diagram I select only those people with the 37 FTDNA markers
from John McEwan's table of those who have been SNP tested. This in itself
makes it less likely for the program to find a suitable link for individuals
with rarer haplotypes. I then remove CDYa & b as these really seem to
create some strange patterns of links and connections due to their high
mutation rate. However, I leave in DYS464 a, b, c, d as this seems to help
to stabilise the diagram - if they are left out, again, some very strange
patterns are produced.

The bias towards samples from those of British ancestry necessarily hampers
anyone from any other origin trying to find connections where there may not
necessarily be any. The program is using this limited and biased data
through a series of passes to try to connect each individual into the
diagram. It seems to be able to some extent to identify clusters such as
"NW Irish", "Frisian" and "Scottish" varieties however, I am in the same
situation as you, where each draft of the diagram will connect you to
different individuals as the data which makes up the whole data set changes.
This is symptomatic of the program reaching the limit of its capabilities, I
believe, as it struggles to find suitable connections. Other individuals in
the diagram will be more fortunate where their links to certain individuals
will remain even though that cluster may be plotted in a different part of
the diagram in subsequent redrafts.

In this instance I haven't weighted the markers in order to produce the
diagram - I have previously done this and again it has produced some rather
weird results. It would therefore appear likely that the diagram could well
be making its connection in your instance based on DYS385ab if this is
indeed the only connection between yourself and the other individuals.

I suppose it is probably plotting relationships which exist but also
estimating and plotting those which may not exist. The knack is to do as
you have done and to determine whether or not you think the relationship
shown makes any sense given what you know about your own family history and
your history related to others.

If you move in close using the zoom facility you will see that actually the
branch you are on comes directly from a spur from the Mode for the whole
group (the yellow circle) and not from the Welsh Modal which is the green
dot to the left of it. This spur leads off to about 1/4 of all the lines on
the diagram. Your particular branch is made up 3 sub branches. Lennan
(Irish) NM35D which goes off on its own. The branch you mention which
includes you but also includes Ozment (S21+) & Dilks both from England. The
third branch is comprised of Evans and Morgan (both Wales) and Kenyon
(England). This is a real mix of individuals and so I believe that nothing
conclusive can be drawn from this until such time as more data from more
individuals is plotted to hopefully make properly identifiable subgroups
which stand out on the diagram.

John Eckersley

This thread: