This sounds reasonable to me. I'd like to hear what folks at the L513 Yahoo Forum would have to say about it.
toggle quoted messageShow quoted text
On 3/20/2017 10:52 AM, Jared Smith wrote:
Aging our ancestors using Y-DNA data is far from an exact science. I'd
be happy to have you poke holes in any of this.
An analysis of the 11 Z16357 people who have taken Big-Y results in
the following number of 'good', unique/novel variants/mutations:
C. Hays 4
R. Hays 3
M. Hartley 5
J. Hartley 5
J. Smith 12
These are variants that each person has that are not shared with
anyone else who has tested. The higher the number of novel variants,
the further back one would expect to be related to someone else
listed. I use the same metric for a 'good' variant as Alex does on his
Big Tree. This is a bit more aggressive than what YFull uses.
There are, however, some inconsistencies with this. Merrick, for
example, has nearly twice as many novel variants as Thomas, even
though Merrick connects lower/later on the tree than Thomas - one
would thus expect Merrick to have fewer novel variants. This is
primarily a factor of test coverage, but this is all we have to work
with, so we partially account for this variability by averaging. This
is why each new Big-Y test gives us increased accuracy.
When I add the novel variants above to the number of 'good' SNPs in
each block or haplogroup of our tree and average the results, I end up
with the following average number of variants downstream from each
listed SNP block:
ZS349 - 3.5
Z16854 - 9.3
BY15420 - 8.0
BY15419 - 9.7
Y29969 - 9.5
A11132 - 5
Z17911 - 10.9
Z16343 - 13
Z16357 - 36.1
This means, for example, that there's an average of 3.5 variants that
were formed after the most recent ZS349 ancestor that the two Hays men
share. For Z17911, we average 10.9 variants downstream (more recent
than) our most recent common Z17911 ancestor. Altogether, we average
36.1 SNPs downstream of Z16357.
To use these variant numbers to help us in aging, we need to calculate
a "years per SNP" value. YFull has our last Z16357 ancestor at around
3300 years ago (though they've acknowledged this is probably too
high). Other recent estimates put it as young as 2300 years ago. Until
someone digs up some Z16357 remains or we get enough DNA testers to
give us better data, we have to use our best informed estimate. I'll
assume our most recent Z16357 ancestor lived a minimum of 2500 and
maximum of 3000 years ago.
If we divide these age estimates by 36.1 SNPs (on average), this is a
minimum of 69.3 years per SNP and a maximum of 83.1 years per SNP. We
can then use these values to assign age estimates to notable
branchings as follows:
ZS349 - 327-376 years before present
Z16854 - 732-861
BY15420 - 639-750
BY15419 - 755-889
Y29969 - 743-875
A11132 - 431-501
Z17911 - 837-987
Z16343 - 986-1166
Z16357 - 2585-3085
The values are years before present, and include an additional 35
years (one generation?) to account for the age of the last ancestor
that had this SNP - and also adds 50 years as a guessed average of how
old the 11 Z16357 people are.
So this estimates that the common ZS349 ancestor for Hays was born
347-376 years ago. We know this ancestor was George Hays who was born
in 1655 - 362 years ago, so these numbers align perfectly!
This places our Z17911 ancestor being born between 837 and 987 years
ago. It places the Hartley common ancestor between 431 and 501 years
ago, the Bennett/Phillips ancestor 639-750 years ago, etc.
Do keep in mind that accuracy is more variable near the end of the
branches (closer to present day), especially with data from only 2 or
3 people. And SNPs are not always formed at a consistent rate. So this
all a bit rough, but should give us fairly reasonable estimations.