GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2006-09 > 1157118815


From:
Subject: Re: [GENEALOGY-DNA] can we do better for genetic distance?
Date: Fri, 01 Sep 2006 13:53:35 +0000


First, I do not claim credit for an equation that was in the paper by Walsh. He used an "infinite allele model". Second, I am an engineer and an approximation for a number that is to be used only as a comparative guide should be good enough. My past dealings with theoretical mathematicians has taught me that they can never design an airplane or do carpentry as they are never satisfied with the sufficiency at the boundaries. :) (My old slide rule tells me tht 2 times 2 is 3.9999.. which is close enough to be 4 which is an old joke told to me by my father who claimed that he passed physics only because the professor, some bridge designer named Wien, passed all pre-med students automatically.) sam

<< 01 Sep 2006 From: "Alfred A. Aburto Jr."

Assuming 1 step differences is a start of course and the idea below
would still give one a better estimate than genetic distance alone.

I wonder how often multiple step differences occur? Any ideas?
If they do occur often then what could one do to handle them in a simple
algorithm? Al

> John Chandler wrote:

>Sam wrote:
>
>>May I suggest
>>
>>Most_Likely_Estimate =
1/(2*Mutation_Rate))*LN(Number_of_Markers/(Number_of_Markers-Number_of_Mutations
)) = 31 generations given 25 matches out of 32 markers and a mutation rate of 4
per 1000 events

>
>You seem to have spelled everything out except "LN" -- did you perhaps
>mean "ln" instead? If so, then that formula is indeed a good
>approximation for the MLE, but only if all the marker differences are
>one step each. A two-step difference shifts the MLE more than a one-
>step difference, but less than two one-step differences (provided you
>recognize that multi-step mutations do occur).
John Chandler


This thread: