DNA-R1B1C7-L Archives

Archiver > DNA-R1B1C7 > 2008-02 > 1203263728


From: "Ken Nordtvedt" <>
Subject: [DNA-R1B1C7] Fw: Age of R1b1c7
Date: Sun, 17 Feb 2008 08:55:28 -0700


My thoughts are enclosed in [[[[[[ ]]]]]]] Ken

[[[[ Bottom line: stick to the variance method rather than GD method ]]]]]
----- Original Message -----
From: David Ewing
To:
Cc: John McEwan ; Ken Nordtvedt
Sent: Saturday, February 16, 2008 10:31 PM
Subject: Re: [DNA-R1B1C7] Age of R1b1c7


In this thread we have been discussing how we might use the increasingly large number of R1b1c7 37-marker haplotypes available to estimate the age of R1b1c7. I don't pretend to have the answer to this, but thought it might be useful to share my thinking about the problem with the list--I would welcome having any fuzzy logic or mistaken notions pointed out to me. I have copied this posting also to John McEwan and Ken Nordvedt, both of whom have considerable experience in these matters, in hopes that they would be willing to instruct us in this a bit. [[[[ Leave out multi-copy markers such as 464 where interpretative work is sometimes needed to count steps of mutation ]]]]]

The average of Chandler's (http://www.jogg.info/22/Chandler.pdf) estimated mutation rates for the 37-marker FtDNA panel is 0.00492. If we eliminate the two CDY markers from consideration, the 35-marker average is 0.00318. [And if we eliminate all five markers with mutation rates >.008, the average rate for the remaining 32 is .00265. The average rate for the slowest 23 markers (those with rates <.004) is 0.00148. I have thought that this kind of calculation should somehow favor slower moving markers, but Ken Nordtvedt has argued that this is a mistake. [[[[ I doubt if I would have suggested favoring fast mutators. I certainly leave them out of my GD work because of the back mutation issue. Of course, using slow markers loses discriminatory power and leads to larger variance of the GDs statistically KN ]]]]]]] Sadly, I can't remember his rationale, though I found it convincing at the time.]

For the sake of argument, let us accept a date of 450 AD for Nial of the Nine Hostages, which is 1558 years ago. At 30 years per generation, this is 52 generations. Now, I understand that predicting mutations in the usual way is fraught with some peril when using time intervals this large (is this because back mutation becomes a significant factor, or for some other reason--genetic drift, perhaps?), but let us see what happens if we use Ann Turner's mutation calculator. You can download this from http://members.aol.com/dnafiler/MutationCalculator.exe
or if you are afraid of the .exe file, you can read about it first at
http://archiver.rootsweb.com/th/read/genealogy-dna/2002-11/1037233627.

Enter 35 markers, 0.0032 mutation rate and 52 transmission events. This yields 5.8 "expected mutations," distributed (Poisson distribution) roughly as follows: 0.3% of Nial's 50th great-grandsons will have zero mutations, 1.7% will have one mutation, 5% will have two, 9.7% will have three, 14.2% will have four, 16.5% will have five, 16% will have six, 13.3% will have seven, 9.7% will have eight, 6.3% will have nine, 3.7% will have ten, 1.9% will have eleven, less than one percent will have twelve, etc, etc. So "on average" his descendants will be genetic distance six or seven from the modal, and those of his 50th great-grandsons for whom he is the most recent common ancestor should be at an average genetic distance of fifteen or so from one another (because two such great-grandsons would be separated by 104 transmission events). Remember, I am leaving CDYa/b out of consideration, here. [[[[[ There is not independence of all the pairs even if you eliminate obvious family "duplicates". The population has been structured all the way back to the NW Irish founder. A large fraction of the pairs of present day haplotypes have their MRCA more recent than the haplogroup founder. So the distribution of GDs that one should expect is a melding together of distributions for MRCAs more recent or equal to the ultimate founder. This composite distribution depends on the demographic history of the haplogroup. ]]]]]]]]]

[[[[[[[ Why not stick to the population variance --- both variance with respect to the founding (modal) haplotype of NW Irish R1b and the variance with respect to the population average. Only one variance need be determined; the other is a simple correction factor from the first. These two variances bracket the age of the haplogroup since its MRCA. I believe you will find the MRCA for NW Irish R1b to be way before the Niall guy; at least I did last time I estimated this age. ]]]]]]







My impression is that the "diversity" of R1b1c7 is not this large, though I don't know of any data that systematically analyzes FTDNA 37-marker panels, excluding the CDY markers. Does anyone have a simple way to tablulate the genetic distances between a large number of R1b1c7 haplotypes or to calculate an average of such? One potential problem in taking this approach would be to include too many close relatives, because this would understate the average difference. Maybe we could get around that by just using R1b1c7 surname modal haplotypes rather than considering all the individual haplotypes we can find. [[[[[ Because of the clan nature of Irish and Scot surnames, and presence of many closely related haplotypes in the databases due to clan projects, I recommend taking just a single example from each surname. Either select randomly or use the family modal haplotype. But if you use the latter, remember you are moving to a haplotype that existed some number of generations ago. This may not be too big a problem. ]]]]]




David Ewing



This thread: