Control Charts for Maximum Values, Not Means
- September 12, 2017 at 2:39 am #55818
I have a knotty problem in a process improvement team – one I’ve not been able to fully understand and explain.
The tool owner is measuring defects on a subgroup size of ~6 (variable) – I’ve suggested a U-Chart does what he needs – so far so good.
He also wants to record and run a control chart for the MAX count value from the subgroup – Using an XBar chart. I’m pretty sure that a) The distribution of X(max) isn’t normal, and b) A correctly configured U-chart will do away with the need for it – especially with n=6.
Thanks in advance!
NSeptember 12, 2017 at 6:03 am #201827
Ask the question 1. What variation are you trying to understand? It seems he’s worried about too much subgroup variation–so why not do a Rbar chart. 2. If the max really is important, would a poke a yoke be better if doable?
If they know what action plan would exist if the max goes out of control for the Xbar chart, then I’d say try it. However, are they concerned if the max gets too low? I’d suggest a run chart possibly.
You must be logged in to reply to this topic.