GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2008-12 > 1228432503-01


From: "Ken Nordtvedt" <>
Subject: Re: [DNA] Tim Janzen's variance calculator results
Date: Thu, 4 Dec 2008 16:15:03 -0700
References: <936471110812041458x72e5e29dv33fdfb7ce842ebff@mail.gmail.com>


As I suggested, I believe he and I combine the variances of the many (N)
markers of haplotypes differently. This certainly will result in different
confidence intervals for the two methods. If the mutation model were as
simple as the basic model, then the two methods should have the same
expected values of their estimators. But I am not sure how the central
limit theorem works on a method which does not combine the separate marker
estimators in a way different from the optimal method based on the variance
of each marker variance. I'll look into that.

Hopefully Heald or Janzen will remind us of the Heald formula. I think the
two are:

G* = Sum of { Var(i) / m(i) } / N (Heald)

G* = Sum of { Var(i) w(i) } / Sum of {( m(i) w(i) } (Ken N)

with w(i) = 1 / (1 + m(i) K ) with K a constant determined by properties of
the tree (mainly G, itself)

Ken






----- Original Message -----
From: "Victor Villarreal" <>
To: <>
Sent: Thursday, December 04, 2008 3:58 PM
Subject: Re: [DNA] Tim Janzen's variance calculator results


> >
>>
>> [[[[ Why would you expect different formulas or "methods" to yield
>> consistent results? I believe his method is incorrect (at least I did
>> the
>> last time we were debating this), and he no doubt thinks his method is
>> correct.
>>
>> Ken ]]]]]
>>
>
> If both methods are based on alleles' variance (and the same Chandler's
> mutation rates) is it not logic to expect some consistency in the output
> differences.
>
> Victor
>
> -------------------------------
> To unsubscribe from the list, please send an email to
> with the word 'unsubscribe' without the
> quotes in the subject and the body of the message
>


This thread: