GENEALOGY-DNA-L Archives
Archiver > GENEALOGY-DNA > 2005-01 > 1105671756
From: (John Chandler)
Subject: Re: [DNA] Male Line Specific Y-STR Average Mutation Rates -- the one size shoe/(haplotype mutation rate) fits all approach is not valid, imo -- corrected URL link
Date: Thu, 13 Jan 2005 22:02:47 -0500 (EST)
References: <LAEMICGGMBKIADLLMIJEKEBEEMAA.terry@bartonsite.org>
In-Reply-To: <LAEMICGGMBKIADLLMIJEKEBEEMAA.terry@bartonsite.org>
Terry wrote:
> My original calculation used the same logic of mutation opportunities and
> actual mutations as the recent discussions (counting as one mutation the
> shared result of several men and only those generations of mutation
> opportunity). The estimate comes from the fact that I had to make an
> assumption of how many generations were missing where the paper trails don't
> connect. (I have some 25 separate families in this group of men - as
> defined by earliest known ancestor). If one assumes a nominally quick
> merging of the 25 families into one family, at 26 markers, one gets the
> .0016 mutation rate. If the common ancestor(s) lived longer ago than I
> assumed, then the calculated mutation rate has to go down further due to
> more mutation opportunities.
The first thing to note is that the average mutation rates are different
for different panels of markers. Your rate of 0.0016 +/- ??? should be
compared against other results for the 26-marker panel, not against the
quite different 25-marker FTDNA panel. I've heard reports that the
RelGen average is lower, and your report lends support to that. What
we need, perhaps, is to settle on a "fudge factor" to convert each
available panel to a common reference.
On the other hand, I'd say that guessing generations is not a good
basis for calculating rates. What happens if you skip the unification
effort and just count the mutations and transmission events within the
documented portions of the tree (i.e., separately within each of the
25 families)? You can then add up the counts for the separate
families and get a more defensible result.
> A while back, I calculated that this group of men were tested on an average
> of 31 markers. When I get the 18 results back, the average will be closer
> to 35 or 36. How many men with 35 (or 37) markers, with each contributing
> an average of 5 generations of mutation possibility would it take to have a
> "statistically valid sample"?
It's not a question of valid vs invalid. No matter how many men you
have, you need to do two things: (1) calculate the statistical
uncertainty in your observed rate, and (2) either compare your numbers
only against other projects with the same mix of markers or else find
a way to correct for the mix.
John Chandler
This thread: