Hello all, after taking over an MSA I got zero in NDC (distinct categories). I suspect it is because the spread in parts is too low.
Can someone cast light a bit of light on the data and tell me whether it can still be useful to assess whether the method is useful? See the attachment.
The L category value determines a different test set, so data should be grouped by L.
Can do something to improve, like augment data with new runs on different levels, for instance? I get that NDC=1.44xPV/GRR, so I guess only PV can be changed, but can it really get to where the NDS is high enough, like 5?
I'd hate to redo experiments, because it has taken quite a long time to get to where we are I'm told.
The test is nested (only 1 part per test), and the spec limits are here, per L value. I think we're using k=4 as sigma multiplier. And as I said, grouping is by L (per the different spec limits):
L0: 1.9-2.3
L1: 10.8-14.0
L2: 4.23-5.53
L3: 18.3-24.3
L4: 24.7-38.7
