Archiver > GENEALOGY-DNA > 2008-12 > 1228160427

From: "Ken Nordtvedt" <>
Subject: Re: [DNA] TRMCA for R1b1
Date: Mon, 1 Dec 2008 12:40:27 -0700
References: <><>

----- Original Message -----
From: "Doug McDonald" <>

> The problem with STR use is the possibility of saturation in variance:
> that is, eventually, the postulated increase in the ratio
> downward rate
> ________________
> upward rate
> as the length increases will cause a fixed variance which will
> not further increase with time. If this is not taken into account,
> times will be too short. There is actual evidence in the
> data that has happened with faster markers, and I don't just mean CDY.
> I think that one should try extrapolating to only the slowest markers.

This saturation of variance which could, indeed probably, exist(s) will
render the variance/age relationship non-linear, so that the underage bias
as percentage of age estimate grows with age of what node is being
estimated. If you look at the number of mutations from founding values for
the clades of 4000 years or so in age, there is no chance yet for much error
due to the saturation effect. But I agree that one should down weight or
eliminate the fast markers as much as possible, but remembering that this
really blows up the statistical confidence intervals rapidly, so it it not a
costless remedy.

This thread: