# daily six sigma quality tip

Six Sigma – iSixSigma Forums Old Forums General daily six sigma quality tip

Viewing 9 posts - 1 through 9 (of 9 total)
• Author
Posts
• #34731

Statman
Member

In a recent daily six sigma quality tip, two pizza parlours preparation times were compared using either range or standard deviation. Standard deviation was preferred, but the conclusion was that there was no difference between the two parlours. Surely, a simple X mR control chart would demonstrate that both were in control, and that one was far more consistent, ie less variation, ie different. Why not choose a control chart EVERY time to compare two comparable processes ?

0
#96780

Jonathon L. Andell
Participant

Could you share the source of the article? I’m curious…

0
#96781

melvin
Participant
#96819

Jonathon L. Andell
Participant

Thanks for the reference. I interpet the message as being: “mean alone is insufficient information.” Those who restrict their approach to comparing means would mistakenly see no difference, but variation is in fact significantly different.
I think that was what the author was trying to convey.
By the way: the data appear to have been sorted. If those results had occurred randomly, the patterns would be alarming.

0
#96849

Ken Feldman
Participant

Jonathon, bet you thought they lifted the pizza thing from your book?  Guess it was better than the seafood restaurant.

0
#96850

Savage
Participant

So what’s the answer to Statman’s question?
Why not choose a control chart EVERY time to compare two comparable processes ?
I don’t know, but sure am curious.

0
#96853

Keith
Participant

In most cases you can, though it depends on the assumptions and criteria used in building the applicable control chart (i.e. – how control limit were calculated, whether the underlying statistical distribution justifies the control chart used, etc.)
In certain circumstances it will be hard to visually distinguish statistically significant shifts in mean or standard deviation (for varying levels of significance). As a follow up, a Z, t, F, chi-square (and so on) test may be needed to quantify the answer.

0
#96854

Jonathon L. Andell
Participant

I probably would consider including control charts just about every time. I’d want to be confident that any difference I see is due to the “X” being evaluated, but that applies to every kind of statistical evaluation. As long as software makes it so easy, I might also look at box plots, ANOVA, dot plots, and other displays –  especially if I was trying to find the most crystal clear way to communicate to management. With such ease, and so many tools, we’d be crazy to get dogmatic.

0
#97156

Bhargava
Participant

Agree with the majority of responses above. You might use a simple t or f test to look at differences between the two parlours.
The use of a control chart would be great to use as it also lets you compare the two process movement on the range chart, which essentially quantifies, how the process behaves in terms of its consistency in delivery time for the very next order. Thus it tell you how stable and consistent your process is.