Date: Sat, 6 Jan 2001 12:08:37 -0600
From: David Hussar
To: Greg Gogates
Subject: RE: Uncertainty
I'm struggling with a lot of these questions myself, but here is how I
understand the logic.
Given your example and assuming the uncertainty estimate was derived with
K=2, there would be a 95% probability the measurement made using your device
would be between 2.88969 and 2.88981 (2.88975+-.000060). Given this range
and assuming a normal distribution it is unlikely that the sample is in
tolerance. In order to say with the same degree of confidence that this
sample is in tolerance you should subtract the uncertainty from the
specified tolerance range. Instead of testing to a tolerance of 2.8894-
2.8897 you should only pass the item if your readings are between 2.88946
and 2.88964.
However, there is more to consider. The uncertainty given only relates to
the calibration of the gauge not to the uncertainty of your measurement. You
would need to determine the other factors of error and their corresponding
uncertainties.
(am I starting to get it?)