– You can have an error in the calculation.
– There can be an error in the specifications (like stating +/-2 when it should have been +/-0.02)
– As Rams said, you can have a process that has a variation that is smaller than the resolution of the measurements. For example, in a good turning process, using an instrument with a resolution of 0.1mm will tyipically deliver R=0 for most subgroups, because the process variation is in the order of a few cents of mm that the instrument can not detect. Note that, even when you get a R=0, the “actual” range of the parts is not (can never be) zero because there IS a difference between the parts, only that you can not detect it. Having a lot of “fake” R=0 will deliver a “fake” too low Rbar. Do you have a lot of ranges = 0 in your data set?
– As Bess said, you can have a process that is a lot more “precise” than what is needed. Sometimes, this means that you are wasting money because you could be using a less precise process but that is still precise enough and at a lower cost (less investment, less mantainance, lower cicle time, etc.).
So, tell us. Which is your case?
.
