A measurement control program to meet desired levels of precision and accuracy

Measurement control programs are usually designed to test for precision and accuracy. Many instruments, however, display non-random data patterns such as biases or seasonal variations which are statistically significant but are of no practical significance. Application of the usual statistical tests can cause these instruments to be removed from service unnecessarily. It is tempting to try to overcome this problem by frequently adjusting the instrument or by arbitrarily changing the parameters of the statistical tests so that failures occur less often. This, of course invalidates the statistical tests. In the author's opinion, the correct way to handle this problem is to identify the desired levels of precision and accuracy, and then to combine these levels with valid statistical techniques in the measurement control-program. This paper describes one way to accomplish this