GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2009-12 > 1260828868

From: "Ken Nordtvedt" <>
Subject: Re: [DNA] R-U152 and R-L21 on the European Continent
Date: Mon, 14 Dec 2009 15:14:28 -0700
References: <200912140842.nBE8gmwK001222@mail.rootsweb.com><6C902AF4FF72438E97B0BC46DF3B709C@john>

----- Original Message -----
From: "Alister John Marsh" <>
> In every case above, the short end confidence interval is able to be
> proved
> to be impossible. Perhaps this shows us that the maths does not yet
> reflect
> reality, at least with respect to confidence intervals.

I think the sensible way to quote confidence intervals is as follows:

Let's assume 95 percent bands are your thing, then:

97.5 percent point of G(low) variance distribution = G(estimated) defines
G(low)
2.5 percent point of G(high) variance distribution = G(estimated) defines
G(high)

G(low) to G(high) is your confidence interval to go along with the
G(estimated) obtained from the variance(observed).

G(low) will never go into the future.

For variance distribution sigmas substantially smaller than expected value
of variance, the above reduces to the method which has typically been used
by people. Ken