On 24 Nov 2003 11:29:40 -0800, [EMAIL PROTECTED] (Find_Housing) wrote:

> Thank you Rich!
> That's a very good point.
> 
> Let's say, if 10 bacteria is acceptable for a bag of 50 ml blood. How
> would you approach the problem?

If you are going to talk about a number so close to zero, 
I think I would approach it from the side of, 
"What can I  believe, if the sample looks perfect?" 

Here is an approximate method, and I don't know 
whether you want 1-tailed or 2-tailed testing.

Consider the 2x2  contingency table, and the 
chisquared  test on the table.  In one row, place the
count of Bacteria 'observed' in testing, and in the other 
row,  place the number which may exist, allowably but
at the limit, in the untested part.    Then in the second
*column*,   place two numbers which are large numbers,
because they are to indicate the fixed-expectations, 
based on the fraction of the sample to be tested.

For instance, if you spotted 0  bacteria, in a 10% sample:
and hope for no more than 10  in the rest --
0   for 1000
10 for 9000 ....  
Those yield a chi squared test value of 1.1; totally ordinary.  
You might check out the fact that the table has almost
the same test value if you change the 1000/9000 to ten
times as much, or (not quite as firmly) ten times as few.
(I'm showing how to do computation with a computer program
that is set up for the numerical entries.  If you have one that
takes fixed proportions, then you can merely put in those
fractions.)

For the table to be 'unusual', you need to escalate to

0 for 1000
35 for 9000 ...
Those yield a chi square of 3.87 (p=.049) or 2.81 (p=.093)
depending on whether you figure Yates's correction applies.
But that indicates that '35'  is in the ballpark of what you can
have some confidence in, for a 10%  sampling.

If you want more assurance based on a zero count, you have
to increase the size that is tested.  If you sample, say, 20%

0 for 2000
20 for 8000 ...
Those yield a chi squared of 4.97 or  3.78 -- so the *bigger*
fraction involved in the testing shows that there must be a 
lower rate, plus, there is a smaller quantity (only 80%)  where
the other bacteria can reside.  

Does this do it?

-- 
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
"Taxes are the price we pay for civilization." 
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to