Advanced MSA of Continuous Data Part 4: How to Sample Parts

Agreed. Operators should be representative of those expected to use the devices, not engineers or people especially proficient in their use.

If you use the historical process variation approach, stacking the deck with parts will have no effect. It is only when you use the parts in the sample to represent the process variation (which is typically impossible to do, anyway) will it affect the outcome. An observant QE can back-calculate your Ppk from the sample you used, and if the sample represents a process that is not capable, then you need to go back to the drawing board and fix your process...or use a truthful sample.

The use of skilled operators is a whole different lecture....but the short story is you should use 1 proficient operator, 1 typical operator, and 1 very new operator or maybe even a temporary worker. That will let you know if the gage system (not the gage itself) is operator dependent in its capability. Very useful. It will also prevent the need for annual gage R&Rs. There is NO TIME FUNCTION in gage R&R. Unlike stability and calibration, once a gage R&R is performed, it answers the question "is the gage the correct gage for the job?", and you are done. Annual gage R&Rs are an abject waste, unless you did them with only skilled workers, and now the gage is used by temps. Do it right, and it is good forever. Your customers may not know that....but there a lot of things they don't really know.