GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2008-06 > 1213116353


From: "Ken Nordtvedt" <>
Subject: Re: [DNA] mushing
Date: Tue, 10 Jun 2008 10:45:53 -0600
References: <200806100823.m5A8NigD018730@mail.rootsweb.com>


I give you a cooking wind up timer clock, a $19.95 timex, and a really good
time piece. I ask you to time me in a 100 meter dash (which I would now be
nuts to try given my arthritis). You need a way to combine the readings of
those three clocks to come up with my time. Afterall, I ran the 100 meters
in one unknown time --- not three. You have to weight somehow the three
clock times. Even if you throw out the cooking clock time and the timex
time, that is a weighting, albeit extreme. Your haplotype has many markers.
Each marker by itself could be used to come up with a GD time for a pair of
haplotypes or a variance time if you have a larger population of haplotypes.
You have to use some criterion for weighting these various marker clocks ---
you would be foolish to judge them equal and take the simple average ---
although some have proposed doing that on this list. DYS570 may show many
mutations (ticks) of its clock; DYS426 may show no mutations (ticks).


----- Original Message -----
From: "Elizabeth O'Donoghue" <>


> Ken, your explanation of weighting helped me a great deal - and without
> even
> one equation in it! Much appreciated.
>
> Does it mean that since all the markers used have different speeds of
> mutation, the weighting in used in order to arrive at an average mutation
> rate when calculating against the total genetic distance between two (or
> multiple) haplotypes? [[[ Darn those folks at FTDNA for dragging in
> "average mutation rates". The operative quantity that divides total GD is
> "sum of marker mutation rates". It so happens that sum of rates is the
> same as average rate times number of markers, so someone thought it cute
> to use the latter formulation --- big conceptual mistake. Weighting is
> necessary in order to intelligently combine the individual marker age
> estimates (which will disagree with each other) into a single age
> estimate. The traditional "flat" weighting is a weighting, nevertheless.
> It is not the absolute best, but as long as your clade is not too ancient,
> it is not bad. See the new table Jim Cullen and I added to our websites.
> Weighting as I have talked about it is for combining marker variances in
> age estimate formulas. The story is a bit different if you are using GDs
> for age estimates.]]]]


Does that take into account whether a genetic
> distance of one, for instance, between haplotypes is a slow mutating
> marker
> rather than a fast marker? [[[[ Moving to GD which although being a close
> cousin of variance, is a different thing; in lowest approximation you are
> simply summing the probabilities of any marker mutating and that is why
> the denominator is the sum of marker mutation rates, while the numerator
> is the sum of all marker changes that occured. This simple GD expression
> becomes more complicated for estimating very ancient GDs, but genealogists
> don't hit this domain. Clearly the fast markers add more to that
> denominator sum of marker mutation rates, and on the other hand they are
> more likely to be contributing to the numerator sum of seen changes.
> Please take "average" out of your thoughts; things are not being
> averaged --- darn FTDNA! ]]]]]
>
> I know that FTDNA's TiP calculator will provide different generational
> distances between different pairs of haplotypes which may have the same
> genetic distance but mismatches at different markers, so it must take into
> account the relative mutation rate of each marker individually. [[[ If
> TIP is doing that they are being too cute given the reality of the
> applications. Have you noticed how small such differences are? ]]]]]] I
> know I've
> read on the list, perhaps even from yourself, that the different mutation
> rates between various markers in the selection used make little difference
> to overall TMRCA [[[[as long as the sum of rates is the same ]]]]], so
> that average mutation rates are used in the
> calculations [[[[ No, I would never say that; and if I did by mistake talk
> about "average" I should wash my mouth out with soap ]]]]]; but I would
> think that the longer the time distance between
> haplotypes, the more likely it would be to make a difference in the
> result. [[[[Yes, for applying GD to ancient TMRCAs, and I really mean
> ancient, complicated corrections depending on the individual mutation
> rates are required. ]]]]]
>
>
> Thanks again. Elizabeth
>
>
> Internal Virus Database is out-of-date.
> Checked by AVG.
> Version: 7.5.524 / Virus Database: 269.24.1/1469 - Release Date:
> 27/05/2008
> 13:25
>
>
>
> -------------------------------
> To unsubscribe from the list, please send an email to
> with the word 'unsubscribe' without the
> quotes in the subject and the body of the message
>



This thread: