GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2011-11 > 1322544355


From: Terry <>
Subject: Re: [DNA] Sigma for Variance
Date: Tue, 29 Nov 2011 16:25:55 +1100
References: <F20455F232334D5B82DD23AD281CB093@kenPC><CACPdyD=LPOcxVteUMB9Jd6OP0v0e3QUWh=LLCJ+K6iheC2cPWQ@mail.gmail.com><144830EBD5E148A99B49B2AAB4DB8200@kenPC>
In-Reply-To: <144830EBD5E148A99B49B2AAB4DB8200@kenPC>


Ken,
Your formula is this:
sigma^2 = m*G*(1+m*(2*G-3)) - m*Gstar*(1+m*(4*G-3)) +
2*m^2/N^2 * sum(G(a,b)^2, over a =!= b),
where Gstar and G(a,b) are things you define, and where (as you mention in
your reply) your formula should additionally be divided by the number of
STR markers (N_STR) if one is combining dates from many independent markers.

Perhaps, the formula should instead be this?:
sigma^2 =(approx) 2/N_STR * (2*m*G)^2,
where N_STR is the number of STR markers that are used.

For around 50 STR markers, the later formula would effectively say that the
standard deviation in the expected G (generation number), is about 20% of
the value of that G.

Have you checked your formula for correctness? I agree that some things
are inherently complex, but it is always worth double checking results.

Terry


On Tue, Nov 29, 2011 at 11:56 AM, Kenneth Nordtvedt
<>wrote:

>
> ...
>

This thread: