From: Stan Hilliard
Date: 06 Oct 2001
Time: 15:43:25
Hi Tommy,
YOU SAID: "A company produces x units per yr. They must validate a change in the process. How many units should be built and tested to accurately represent a build of x units. "
MY COMMENT: In my opinion, a process validation shouldn't represent one year or any other time period. Think of a process as a configuration. The data that you collect under that configuration is a sample from an "imaginary" infinite population -- all of the items that would be produced if the configuration were to run forever.
You determine a sample size the same way as for acceptance sampling -- specify the acceptable and unacceptable process levels with points on an oc curve.
Stan Hilliard