Archiver > GENEALOGY-DNA > 2010-02 > 1265781181

From: "Alister John Marsh" <>
Subject: Re: [DNA] TMRCA Estimate of a group of Marsh families
Date: Wed, 10 Feb 2010 18:53:01 +1300
References: <><CA75B5D417594330AF399ACA1F13F2B3@anatoldesktop>
In-Reply-To: <CA75B5D417594330AF399ACA1F13F2B3@anatoldesktop>


Thanks for your further comments.

It is interesting to look at estimates for TMRCA. They can help a person
put "most energy" into investigating evidence from a "most likely" time

Even if the TMRCA estimate is wrong, it draws to attention things which can
be tested. If the TMRCA is eventually proved wrong, then in proving it
wrong (or inconclusive), knowledge has been advanced. I am famous for
making speculations which eventually are proved wrong. If you are afraid of
being proved wrong, then genealogy is not a good hobby to take up!

There are living Tyndalls who have well documented pedigrees back to
Northumberland 900 years ago. It is therefore well within the realm of the
possible that I can reasonably "prove" (or disprove) my Tyndall matches
descend from the de Tyndalls living in Northumberland 900 years ago. If I
could do that, it would give me some concrete facts to refine my
speculations around. Refined speculations might highlight another step to
complete the jigsaw.

In 10 years, no doubt I will be still gnawing away at this genealogical


-----Original Message-----
[mailto:] On Behalf Of Anatole Klyosov
Sent: Wednesday, February 10, 2010 5:58 PM
Cc: Anatole Klyosov
Subject: Re: [DNA] TMRCA Estimate of a group of Marsh families

>From: "Alister John Marsh" <>
>...Perhaps I understand your processes better through you working through
this example.
Your assessment of there being two clusters is correct.
The age you estimate for the larger cluster (A to I) at 400 years ago is
possibly in the right ball park, although one or two of the haplotypes may
be from branches earlier. If anything, the cluster A to I may be a little
older than 400 years.
This smaller group of J and K is only two haplotypes, I suspect your
estimate of 100 years to branch common ancestor is considerably out.
However, because the haplotypes were the same, and only two of them, I think
there is not a large enough group to make predictions with statistical
confidence, which you in effect stated. You can't be criticized for
predictions on such limited information as 2 haplotypes.

My response:

Dear John,

I am glad to be of some help. Perhaps our discussion and a consideration of
your example might be also of some help to people here who elect to focus on
problems and not on their solution. It also shows that there is nothing
mysterious with counting of mutations, that it is good to work with specific
examples and not just to talk generalities, and that it is not a big problem
to dissect a haplotype datasets into separate branches.

I should also add that I like, John, your attitude directed to problem
solving. That is how things are done, not just discussed.

Now, back to your example. As you well know, haplotypes and their mutations
rule. One cannot get a reliable TMRCA with just a few haplotypes and with a
few mutations in them. One cannot beat a high margin of error in those
cases. That was an objective issue in your example. Eleven haplotypes with
two branches in them is not a gift.

What does it mean? It means that statistics is not there. It means than an
extra mutation which can easily happen in any generation shifts an apparent
timespan to a common ancestor. Nevertheless, an estimate can be made.

In your set of 11 haplotypes it was quite obvious that the last two
haplotypes formed a separate branch, aka a lineage. They had a triple sync
deviation from the rest of the pack, and one deviation was by as many as two
mutations in the both haplotypes in the same markers. Things like that do
not happen randomly. It was a system there. To verify it I have composed a
haplotype tree, and - sure enough - those two haplotypes were sticking out
as a sore thumb.

However, those two 37-marker haplotypes contained only one (1) mutation in
all 37 pairs of alleles. It means that a common ancestor for those two
haplotypes lived 1/37/2/0.00243 = 5.6 generations back (to be ridiculously
precise), that is 140 years ago. Since one mutation results in the margin of
error of 100%, we have 140+/-140 years to a common ancestor for those two
haplotypes (in my previous message there was a typo, I was in a hurry).

The rest of 9 haplotypes contained 13 mutations. I have counted that 4-step
mutation as just one on obvious reasons. However, it did not change the
outcome much, as I will show shortly. 13/9/37/0.00243 = 16 generations, that
is 400 years to a common ancestor. A margin of error in this case is 56.4%
for the 95% confidence, that is 400+/-225 years to a common ancestor. If we
count in the 4-step mutation, it will be 16/9/37/0.00243 = 20 generations,
that is 500 years to a common ancestor. As you see, it is within the margin
of error.

Since there are 4 mutations between the both branches, it places their
common ancestor at about 845 years back, that is still at the 12th century.
You can play around with margins of error, which will be certainly more than
a century, however, my experience shows that those margins or error are
typically overestimated. By placing too high margins of error one just fools
himself. I would suggest you to consider - for the time being - that a
common ancestor of the population lived in the middle of the 12th century.
Only if documents contradict this estimate, it can (and need) be
reconsidered. Besides, it can be examined with a more extended series of

>Also I suspect a number of back mutations

No, disregard them. Simple math shows that a contribution of back mutations
is practically zero during the first 26 generations to a common ancestor,
and truly negligible in the first 1200 years or so. "Your" the most recent
common ancestor lived within that range. After about 2500 years to a common
ancestor back mutations are kicking in progressively.


Anatole Klyosov

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: