GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2005-07 > 1120262331


From: John S Walden <>
Subject: Re: [DNA] mtDNA results - Helena
Date: Fri, 1 Jul 2005 19:58:51 -0400
References: <dd.29357684.2ff5f585@aol.com> <fef9eee40507010830292c0ff0@mail.gmail.com> <REME20050701141217@alum.mit.edu> <fef9eee4050701120178bb303b@mail.gmail.com> <REME20050701162656@alum.mit.edu> <fef9eee4050701144729015232@mail.gmail.com> <REME20050701181350@alum.mit.edu>
In-Reply-To: <REME20050701181350@alum.mit.edu>


On 7/1/05, John Chandler <> wrote:
>
> Why should they be close?
>
> John Chandler


Well given all the very small numbers for any thing else
I thnk we are sort of talking about the same thing.
{maybe it is issues like this that kept me from getting a good
grade in graduate school statisics those many eons ago.]

Lets get rid of the times two issue and look at the two things one by one
and let us use just 10 generations
From the Cum Prob. function I get this
g Cum.Prob 1st Diff.
0 0
1 0 0
2 1.18 E-7 1.1E-7
3
4
5 1.17 E-6 5.87E-7
6
7
8
9
10 5.28E-6 1.5E-6

So the culative prob of finding EXACTLY one mutation in 10 generations 5E-6
I would think [maybe this is where I am wrong]
That the prob. of finding anynumber of mutation over zero is also about 5E-6
becasue they are so rare to start out with to find 2 or 3 or ... 1143 sort
of adds in only rare squared [so to speak].

Now if I look at the Poisson calculations
At one generation [or birth]
I get the following results
Zero mutations .99965 [almost one]
1 mutation 3E-4
2 mutations 6E-8
3 and more even smaller

So the dominent value is the 1 mutation at 3E-4

At 10 births I get
Zero mutations .9965
1 mutation 3E-3
2 mutations 6E-6
3 and more even smaller

So this sort of makes sense to me the following way
We have the chance of 3E-7 on the mutation and we have 1.1E4
markers to mutate so the have 3E-4 as the result at one birth.

And if I look at 10 births the odds are about 10 times that of one birth.
There has not been enough time to go by for the odds of a 2nd, 3rd
or 4th mutation to make a difference.

Thus on the Poisson I belive it as it makes sense

And the cumlatave probability fuction it does not make sense
[assuming there is just a when to start counting issue from 0 to 1]
Look at the first step at generation 2
It says the odds of finding one mutation is 1E-7
This is one third the value of having only one marker mutating at 3E-7 each
time.
The Cum Prob does does grow exponentially at the start
but it is to small for each step.


So how should I look at this?

John W

This thread: