GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2011-06 > 1307693069


From: "Sandy Paterson" <>
Subject: Re: [DNA] Asymptotic Distributions for General Mutation Models
Date: Fri, 10 Jun 2011 09:04:29 +0100
References: <972D673E3D084E2DBBD515DD90077C82@kenPC> <000001cc2606$b16669c0$14333d40$@com> <4DEFBCFF.6040501@gmail.com> <000001cc2610$d0ce1240$726a36c0$@com> <4DEFD211.60207@gmail.com> <000001cc26db$a4d9eb20$ee8dc160$@com><4DF16E65.9050205@gmail.com>
In-Reply-To: <4DF16E65.9050205@gmail.com>


You seem to be suggesting that it is wrong to criticise a one-factor model
when that model accurately estimates what can be estimated from that one
factor. I guess that's one way of looking at it.

But I think we need to examine whether IA does indeed give accurate
estimates. For instance, simulation (one-step only) using different mutation
rates for different markers with a mean mutation rate of mu produces very
different results to IA using that same mean mu. Is there a way of varying
the mutation rates by marker in IA? (this is a serious question - I really
don't know IA well enough to be able to work it out).

On the question of the wisdom or otherwise of using just one factor to do
the estimates, I suppose it depends on what you are trying to do. If you
want narrower confidence intervals, the answer is obvious to me. You need
more than just one factor. I'm up to 5 now.

You seem to suggest that this makes a comparison between a multi-factor
approach and IA unfair in some way. I can't agree with you here. The only
way to assess a 5-factor model with factors A,B,C,D & E is to start with A,
and see how the accuracy improves buy adding B, then C and so on. In this
case A = number of matches (or IA).




-----Original Message-----
From:
[mailto:] On Behalf Of David Johnston
Sent: 10 June 2011 02:08
To:
Subject: Re: [DNA] Asymptotic Distributions for General Mutation Models

I am not sure what you mean by diabolically poor. Do you know of
anything better?


This thread: