GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2010-02 > 1266780776


From: "Ken Nordtvedt" <>
Subject: [DNA] Intraclade Sigmas (was S116* ....)
Date: Sun, 21 Feb 2010 12:32:56 -0700
References: <201002211810.o1LIAoGY024816@mail.rootsweb.com>


----- Original Message -----
From: "Tim Janzen" <>


> the point of these intraclade estimates
> is to look for differences in variance by region.

[[[ The problem is that differences in variance by region should be
statistically significant before drawing much in the way of conclusions .
And the quoted sigmas for intraclade age estimates in this tread have been
unrealistically small. One does not gain any 1/N^1/2 factor for intraclade
age sigmas by using a haplotype sample size of N. That factor is relevant
only for independent stocastic variables; the age estimates for the N
different branches to final haplotypes are highly correlated for a pruned
clade tree --- why? Because many of the different branches share a good
portion of their branch paths with each other. And one needs some
assumption of the clade tree structure in any case to determine the sigmas.
I reckon they are about twice as large as one's I have recently seen in this
discussion. Ken ]]]

> More precise conclusions might possibly be drawn at some point in
> the future when we have more data.

[[ One should never oppose more data; but these intraclade variance age
estimates only extremely slowly improve with haplotype sample size. And one
eventually approaches the ultimate limit --- the age sigma if you had the
haplotype of every one of the millions of males today who descend, for
examples, from the I1 MRCA or the U106+ MRCA. We are now often asking
questions about the timing of events within the bushy post-agriculture part
of the y tree. Things are happening so fast, like occurrence of successive
nodes, in these bushy trees I doubt intraclade sigmas can ever resolve
much --- unless of course there are some drastic differences in regional
populations. Ken]]




This thread: