I am testing inventory records for accuracy. How many records do I have to
test before I can say with 98 % confidence that the sample results are the
same as the total would be had I tested them all?
I am testing inventory records for accuracy. How many records do I have to
test before I can say with 98 % confidence that the sample results are the
same as the total would be had I tested them all?
The literal answer should be 98%, however you can perhaps improve your chances by applying a feasibility test, ie, if opening + purchases - sales (= closing) is less than 10% of a months sales, or is more than 20% of a months sales then perform your check on those items.
The standard run-of-the-mill items can usually survive with a minimal check, but in either case the number of items to be checked should be increased for each error found.
Bryan-
Thanks for your suggestion.
I guess I am thinking of the problem more like a process with an expected
value and a deviation of either 0 or 1. I was thinking in terms of
statistical quality control there must be a formula that would allow me to
make inferences about the universe which is finite from a less than 100%
sample.
"Bryan Hessey" wrote:
>
> The literal answer should be 98%, however you can perhaps improve your
> chances by applying a feasibility test, ie, if opening + purchases -
> sales (= closing) is less than 10% of a months sales, or is more than
> 20% of a months sales then perform your check on those items.
>
> The standard run-of-the-mill items can usually survive with a minimal
> check, but in either case the number of items to be checked should be
> increased for each error found.
>
>
> --
> Bryan Hessey
> ------------------------------------------------------------------------
> Bryan Hessey's Profile: http://www.excelforum.com/member.php...o&userid=21059
> View this thread: http://www.excelforum.com/showthread...hreadid=390044
>
>
Based on information in my intro to statistics textbook (I'm not a professional statistician), there are ways of determining sample size for a given confidence limit like you want to do. Exact formulas depend on what type of data you have ("continuous" data or binomial data).
Assuming my stat textbook is representative of all intro to stat texts, you should be able to look this up in a textbook.
There are currently 1 users browsing this thread. (0 members and 1 guests)
Bookmarks