nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Control Chart tests not flagging few values

vs90

Hello,

I have a script which generate IR control chart and tests 1,2,3 & 8 on the data points. When I'm not applying any limits it's working fine:

But when I'm applying limits it's acting weird. Flagging one value and leaving others or not flagging at all. See this:

It's totally out of my mind what can be the possible reason???

Accepted answers

vs90

This was actually a good thing to know when passing limits using variables but not as limit table then not only you have to pass LCL(),AVG,UCL but Sigma() is also important.

Otherwise it'll calculate Sigma based on original data and apply only those three values which causes this issue to arise.

All comments

vs90

This was actually a good thing to know when passing limits using variables but not as limit table then not only you have to pass LCL(),AVG,UCL but Sigma() is also important.

Otherwise it'll calculate Sigma based on original data and apply only those three values which causes this issue to arise.

tonya_mauldin

These tests are based on zones, not the control limits. Click on the red triangle next to Individual Measurement of... and select "Show Zones". See if the tests make more sense now. The Zones are affected by the standard deviation of the control chart.

I suggest using the new Control Chart Builder platform with the "Set Control Limits" option to get what you are looking for.

vs90

Exactly. These tests are based on zones which typically are +3 sigma & -3 sigma, where sigma is standard deviation.

Actually in my script I'm not using "Get Limits" functionality which uses the limit table to apply limits. Instead I'm passing them as variable each for LCL, AVG & UCL.

Now what was happening it was dividing these zones based on standard deviation of original data but not as of applied limits. So the work-around is to pass Sigma() another variable with rest three to get tests work properly.

I have used Control Chart builder as well but in my case there are 3 different charts getting generated for like 12 different processes so used these charts individually only.

Thanks for the suggestion though.

Steven_Moore

While we're talking about signals on control charts (Process Behavior Charts), there are many so-called "rules", but to avoid too many false signals, use the four "Western Electric Rules"; one point outside 3 sigma, 2 out of 3 outside 2 sigma, 4 out of 5 outside 1 sigma, and x in a row on one side of the center line; I generally use x=7, but some folks prefer 8 or 9. If you use these four rules only, you will get better results and live a happier life!

For more details, see Dr. Donald Wheeler's writings.