GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2011-01 > 1294176954


From: "Ken Nordtvedt" <>
Subject: Re: [DNA] How old is Y-Chromosome Adam?
Date: Tue, 4 Jan 2011 14:36:04 -0700
References: <201101040808.p0488xgP020485@mail.rootsweb.com><AA814106-788F-475B-9494-D839E21F7AC6@vizachero.com><AANLkTi=FN3NEtEsp93qpYyLSr_tD9PLJb+pRZm+HHpv7@mail.gmail.com><00e301cbac30$3bb02b80$c2482dae@Ken1><AANLkTikw0bwRedjAOMqFXGJ73jEmh-pXOuLfKQENrB77@mail.gmail.com><010e01cbac34$c3983480$c2482dae@Ken1><AANLkTimhrGmoUa=attiP4TkQxa73QJK9-vVdvoM2free@mail.gmail.com>


----- Original Message -----
From: "Dienekes Pontikos" <>

> The thing to show is that the median/average mutation rate estimate
> ratio decreases with higher mutability.

[[[ It probably increases. Certainly the mostlikely/average ratio does ---
n/(n+1) increases as n (mutability) rises. For median/average ratio, I just
have two points, n = 0 and n = 1. Finding median requires solving
transcendental equation which I have not programmed yet. For n = 0 the
ratio is ln2 / 1; for n = 1 the ratio is solution of transcendental equation
e^u =2(1+u) which comes out to be about 1.7 / 2. I'm confident the ratio
will continue to rise asymptotically to one.

But that's not the real issue. Making the right inference of the very slow
marker mutation rates matters most and significantly in a FRACTIONAL sense
only when you restrict your haplotype markers considered to the slowest.
When you have the mixture of slow and fast markers, what biases you have in
slow marker rates hardly tickles the sum of all marker mutation rates. And
the variances of the very slow markers also contribute little to the sum of
variances. So G predictions are hardly budged by very slow marker biases in
rates in these circumstances. KN]]]


Then you can quantify how much
> of the 10+ fold difference in age estimates between the slowest and
> fastest markers can be attributed to the postulated bias>>

[[[ My message was just a response to a brief question posed by
omeone --- what is the basis for claim that very slow STRs have their rates
systematically underestimated? It was not addressing everything that is
probably going on in the age estimation domain using STRs, especially
anomalies in fast marker, large time variance growth. There is an empirical
approach to isolating the very slow marker rate question. Restrict oneself
to making age estimates for a large number of young nodes in the ballpark of
just a few thousand years.
The model anomalies in the fast markers are now drastically quenched in a
fractional sense. Now do the same very slow, slow, medium, fast, very fast
marker game with these node age estimates that was done with genetic Adam's
age estimate. Any trends still there would more likely be associated with
the very slow rate bias --- if it will even show given the huge statistical
sigmas that result. KN]]]



This thread: