Theo

Why do you want to do this ?  What are you trying to do ?

It is interesting that nowhere on Mario’s ASC site, nor in his papers, does he make any mention of 1.5 or 3.4.  I assume this is Bill Smith and Harry’s fabrication that Mario sees as false ?
Mario does however fall into the trap of talking about “yields of  99.9999998%” – (30 bit precision ! ) – I’d like to see how he measures that in any meaningful way !

A very informative post !
I’d love a copy of the SME Symposium Proceedings
[email protected]
Thank you

Much more importantly, for skewness less than 1 and kurtosis less than 4 (or 4-3 as a departure from normal), 3 sigma control limits on XmR charts give 99% coverage.  In other words, for the majority of situations, you don’t have to worry whether your data is normal in order to use control charts.
This is exactly what Shewhart proposed.

This is a good example of six sigma’s most fundamental flaw, in that it is a specification based methodology. By focussing on defects, or in this case % on time, information about the process is lost.
Can you please provide that actual times.  Before doing that, could you also describe exactly what you are trying to do.  In other words, a p…[Read more]

Time based processes such as this are always non normal. You will usually see a peak to the left and a tail to the right.  Could you possibly post your data please?
If you apply any transform you will also need to apply an inverse transform on any results.  Transforms distort the data and can readily lose it’s meaning.
Fortunately you don’t ne…[Read more]

This is incorrect.  Your data does not have to be normal to use XmR charts.
Read Dr D Wheeler “Normality and the Process Behaviour Chart”. 

Andy,
Yes, the pieces seem to fit.  I assume the 1.5 screw up was followed by Bill Smith and his “brilliant” Cp=2 idea.  Then along comes Harry and mixes up the Cp=2, Pre Control’s green zone, Bender’s stacks of disks and voila, Six SigmaTM.  Later Harry realizes his mistakes but figures he can make lots of money out of it …
I looked up Mari…[Read more]

Fantastic Andy !
At last an explanation that is starting to make sense of all this nonsense. One thing that doesn’t quite fit though; Pre Control’s green zone is set at 50% of the spec limits.  This would give an allowable “shift” in the green zone of  +/- 1.995 , not +/- 1.5 at a Cp=1.33. 
I don’t understand what you mean by “multi-normal di…[Read more]

Correction,
The only paper I’m aware of by Wheeler on Six Sigma is “The Six Sigma Zone”.  As always, it is very well presented and as expected, he explains why six sigma is so much nonsense:
http://www.spcpress.com/ink_pdfs/The%20Final%206%20Sigma%20Zone.pdf
As with the “15-20% sales” thing, he draws attention to the fact that Harry references…[Read more]

Craig, Mike,
What is his phone number and email address ?

Andy,
The 1.5 resulting from confusing Shainin control limits with Shewhart limits is an interesting thought.  The actual origin of 1.5 has never made sense to me. It’s obviously not an observation.  If it was, it would be seen as a special cause.  Once it was created, Harry and Reigle obviously saw it as the basis of a good money spinner (som…[Read more]

Correction,
Well done. Some great investigation and an excellent unveiling of Harry as a fraudster.  Do you know what the field was for his first degree ?  A Bachelor of Arts perhaps ?
Could you post a copy of the article or email it please ?

I assume you are new to the forum here ?
It’s great to have someone who knows what he is talking about.

I think we may be in agreement.  Efforts to fit data to normal distributions in SPC practise, are a waste of time.

RE:  “normality does not exist”. 
Perhaps I should have said “it is impossible for practical purposes to collect enough data to validate that a normal distribution exists for the purpose of six sigma calculations”.
It is also impossible to distinguish, for practical purposes, between a normal distribution and a Burr distribution.

G L
I think you are being a bit harsh on poor Marlon.  Marlon is just parroting more of Mikel Harry’s marketing rubbish.  I seem to recall an article where he claims “4 sigma – 15-20% of sales (Industry average)”.  I don’t suppose anyone has a copy. 
BSI has also obviously been stupid enough to get sucked in.

Sorry Marlon but this is quite incorrect … unless perhaps you have some theoretical justification for your staement “Sigma  level  equal  to  4 should  cost 15-25 % of  the  total  sales”  ?
Please read the dozens of posts here explaining how sigma levels are totally meaningless.
Cp 1.33 is quite arbitrary.  It is a commonly chosen value.  A…[Read more]

You can buy them through SPC press.
http://www.spcpress.com/
The third book “Normality and the Process Behaviour Chart” is quite new … it gives a very entertaining history of statistics, leading to the control chart … an absolutely wonderful read. He includes some surprising revelations, for exampe, why quincunx data is not binom…[Read more]

You are missing the point.  All data is non normal.  Normal distributions are a theoretical construct. We can never know what the true distribution is.  Read Shewhart !
Wheeler proves Shewhart’s original postulation that normality is not required for Shewhart Charts.
Shewhart does not use the CLT for his charts. Wheeler proves it is un…[Read more]