Hello
This is probably a banal question for a full-time statistician, but I
need some advice on the validity of a method I intend to use.
Scenario: A population 1 of purebred dogs in which the chance to reach a
certain age is established fairly well (sample of roughly 400 dogs
during the last 20 years). We have the lifespans of 9 randomly choosed
dogs from an earlier population 2 (between 1900 and 1930. Obtaining
lifespan data from back then is extermely difficult. These nine dogs are
the result of scanning roughly 150 individuals from this timespan).
In Population 1, the chance of reaching 10 years of age is p=0.17, the
cance of reaching 13.5 is q=0.01
Now, population 2 contains both a dog that reached 10 and a dog that
reached 13.5. It should be stressed out that this is not the result of
taking several samples and choosing the most convenient one.
It is easily established that the chance for a 10-year-old in a sample
of nine would be 1-(1-0.17)^9 = 0.81
(in general notation, 1-(1-p)^n)
Likewise, the cance for a 13.5-year-old would be 1-(1-0.01)^9 = 0.086
(1-(1-q)^n)
No big deal.
Alone, neither of the two dogs gives a significant difference, thus it
could well be coincidence. I strongly suspect that having both together
in the sample of nine would make one. However, I have a problem with
calculating the exact chance.
Here is the problem:
For dog a, we have the chance of 1-(1-p)^n
For dog b, we have the chance of 1-(1-q)^(n-1)
But it could equally well be said that
For dog a, we have the chance of 1-(1-p)^(n-1)
For dog b, we have the chance of 1-(1-q)^n
Which gives us two different chances for each dog. Thus, order plays a
role. Now, my problem is: How to calculate the chance without order
having an influence? I hope someone will be able to explain.
Secondly, if the chance of this age distribution in a random sample of 9
is as small as I suspect it to be, would you consider this a valid proof
that population 2 had a higher chance of reaching the ages of 10 and
over than population 1 has?
Thanks for reading
Silvan
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at
http://jse.stat.ncsu.edu/
=================================================================
