Are there any references or standards that dictate the color coding of the zones of a Shewhart style control chart. We use green in +/- 2 sigma range, yellow between 2 and 3 sigma and red beyond 3 sigma. I am hearing that using green, yellow and red is not good because of other connections (like a traffic light) we have to those colors in society. Will you please reference your response? Thanks
$\begingroup$
$\endgroup$
1
-
2$\begingroup$ I don't have too much to say on your exact question, but would advise you not to use red & green together in the same graph. Red-green colorblindness is common in men, ~10%. This page has a lot of good, relevant info. If you scroll down, it lists sets of colors that are unambiguous for all, w/ coordinates for the colors. $\endgroup$– gung - Reinstate MonicaCommented Mar 23, 2012 at 3:34
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
I don't know if there are standards, but the red/yellow/green sounds like a good first try ("warning"/"caution"/"okay") precisely because of traffic lights. As long as that's what you are actually trying to convey. However, the red-green colorblindness issue is a problem.
If you're using R, there is a package called qcc which does Shewhart graphing and I assume they've done some amount of research as to what is the norm. (They appear to use red points to highlight potential outliers.)
You could use colors and shapes: maybe circles for green, triangles for yellow, x's for red, or maybe dots for green, circles for yellow, and circles with x's for red. You could do this with color and shape, or just shape.
You could also use background shading: for example, shade the background of the graph so that it is a solid mid-gray inside of +/- 2 sigma, then a light gray from 2-3 sigma, and white beyond that. Plot black marks on this and contrast will make your point: black on gray stands out less than black on white, so normal data stands out less.
$\begingroup$
$\endgroup$
Don't Shade
I have three quarters of a dozen books readily available dealing with quality control and Shewhart/control charts. None of them recommend using color bands or even routinely plotting zones for most control charts.
The reason is that most modern applications can auto-detect the "8 tests for special causes," which relies on the zones. The most important test, and the purpose of the charts, is to detect truly abnormal behavior or trends.
For the data to be normal, there should be points in every zone. In fact, fifteen points in a row in the "green" zone is one of the flags that your data is not is statistical control. You should have ~4.28% of your data in the "red zones" and ~27.18% of your data in the "yellow zones" when your data is normal.
Sources
The Six Sigma Handbook by Thomas Pyzdek
Implementing Six Sigma by Forrest W. Breyfogle III
Quality Control by Dale Besterfield
Memory Jogger by Goal QPC
Black Belt Memory Jogger by Goal QPC
Tool and Manufacturing Engineers Handbook (TMEH): Desktop Edition edited by Cubberly and Bakerjian
Introduction to Statistical Quality Control by D. C. Montgomery
Six Sigma for Green Belts and Champions by Gitlow and Levine
The Lean Six Sigma Pocket Toolbook by George, Rowlands, Price, and Maxey
