HELP - Sampling Data

[JCombs]

I have multiple types of feed and within each I have multiple analytes. Each feed type had 40 bags. Out of a lot(40 bags), 10 probes (1from each bag) are put together as a representative sample. I am trying to prove that 10 probes has x variability for each of the analytes. In my mind I need to calculate all of the possible different combinations of 10 and look at those means (yes I know its around 8.5 million different combinations).

I cannot figure out how to generate the means to compare in R or any software. Thoughts? ANY help would be greatly appreciated! Once I figure this out for one analyte I should (in theory) be able to carry over the same actions to all of the feed types and each of their anayltes.

 

[obh]

Hi J,

Trying to understand the question...
I assume the feeds are independent? how many feeds?
For one feed you sample 10 probes of the entire population of 40?
What do you check? what is the probability you will find it?

 

[JCombs]

Probability we will find the nutrients labeled 100%. Protein, Fat, Fiber, Calcium, Zinc, Phosphorous, Drugs, etc.
Normally we sample 10 probes (and combine them) out of a single lot of feed to present one sample to the lab. We collected 40 of such probes to analyze separately to figure out the greatest possible spread of any 10 probes. Essentially we want to find out if our "AVs" which are our laboratory AVs plus the prep AV's PLUS a fudge factor - is wide enough to cover the greatest spread possible. And to determine how the means look when compared to each other. We would only look at one feed type and one analyte at a time. Each will be different; however, have a set "violation range" a number above and below that is different for each analyte. e.g. Protein is always 3% above or below before it becomes violative. Essentially we want to determine if our math - that was set in 1950 is still valid and defensible.