Not all characteristics follow a Normal Distribution - How do you do SPC Chart

[amiuda]

Dear Experts,

Not all characteristics follows Normal Distribution:

We can apply some testes like Anderson-Darling test for normality to check whether the data follows a normal distribution.

On the cases the data does not follows normal distribution, how do you perform SPC charts in excel ??

Thanks

 

[Corca]

Re: Not all characteristics follow a Normal Distribution - How do you do SPC Chart ?

Hi,
Not and expert but,
I think increasing your subgroup sample sizes would do the trick because increasing sample size drives the distribution to normal.
You can also work out what your sample size should be.
Then use statistics as normal for analysis.

Wait for an expert to comment to ensure this is the right track.

 

[randomname]

If you're using X-bar chart, don't need to worry since averages will be normally distributed.

 

[Steve Prevette]

SPC works because of the Tchebychev Inequality (which is non-parametric) and due to Dr. Shewhart's empirical testing of SPC against the Normal Distribution, a Triangular Distribution, and a Rectangular Distribution. SPC does NOT require a Normal distribution, NOR is it reliant on the Central Limit Theorm (averaging).

It is amazing the amount of misinformation out there about SPC, a lot contributed by the Six Sigma folks. I would make the recommendation to anyone who wants to understand SPC - Read Dr. Shewhart's original works on the subject.

 

[Bev D]

To add detail to Steve’s response (SPC does not require the Normal Distribution or the Central Limit Theorem to work) you might want to read the following articles by Donald Wheeler. He is a Shewhart/Deming purist and before reading Shewhart’s work you might find Wheeler’s easier to digest and understand.

“Myths About Shewhart’s Control Charts”, SPC Tool Kit column, Quality Digest, September, 1996

“Myths about Process Behavior Charts”, Quality Digest September, 2011 (This has more detail regarding different distributions)

“Foundations of Shewhart’s Charts”, SPC Tool Kit column, Quality Digest, October, 1996


To Steve’s point the amount of misunderstanding regarding SPC – and statistics in general is astounding. Its persistence is distressing. While certainly the hacks that jumped on the Six Sigma bandwagon to make a quick buck spread a lot of this ****, the explosion of the internet and the growth of our desire for “instant pudding” knowledge (as Deming referred to it) as well as the utter lack of humility in trying to truly understand things has really contributed to this.

If you want a bit more detail this is a nice article regarding the debate between Shewhart and Pearson

“Shewhart and the Probability Approach” Quality Digest, November 2015

 

[Chennaiite]

I read couple Donald Wheeler's piece that changed my perception of Control Charts. I began to understand that SPC is a way of thinking and not just an academia-type work-out of numbers.