GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2008-12 > 1228935754


From: "Ken Nordtvedt" <>
Subject: Re: [DNA] Tim Janzen's variance calculator results
Date: Wed, 10 Dec 2008 12:02:34 -0700
References: <200812101850.mBAIoqGj013811@mail.rootsweb.com>


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

I have been concerned about the fact that the
> calculator frequently generates TMRCA estimates that are lower than would
> be
> "expected" almost from the time that I first developed it in May.

Perhaps
> Ken Nordvedt, others, or you can offer suggestions as to how to correct
> for
> this issue using the variance method.

It has been pointed out from the beginning that INTRACLADE self variance
estimates a G* that is less than the G to the MRCA. Some books therefore
describe this as the coalescence time; and a look at the intraclade self
variance explicitly reveals that it is the average TMRCA for the PAIRS of
haplotypes of the sample population. Some pairs clearly have a MRCA who is
a later node in the tree than the earliest MRCA node, and if you pick too
many cousins in your sample population, that average really gets pushed
younger.

This is why I have been advocating pretty much giving up on the intraclade
variances and formulating your search for key times in the tree using the
INTERCLADE variances. INTERCLADE variances do not suffer this particular
problem.

Those who use intraclade variance with respect to an assumed founding
haplotype (and are correct in that assumption) will have an estimate of G
back to that MRCA founder.

Ken


This thread: