Theo

ISO9004 has nothing to do with six sigma.  It is based on TQM’s seven tools of quality.

If the company had previously been practising the principles taught by Deming, ROIC would be negative.

By mixing data you have lost the whole purpose of the control chart.
Read Shewhart, Deming and Wheeler for the basics.

Firstly, what are you trying to achieve ? What problem are you addressing ?
Is your data binomially distributed ?  That is, does it follow the conditions for a binomial distribution, as described in “Advanced Topics in SPC” Wheeler, p255 ?  If not, and assuming you have an adequate sampling technique, and that data is from a single source, use an…[Read more]

Firstly, what are you trying to achieve ? What problem are you addressing ?
Is your data binomially distributed ?  That is, does it follow the conditions for a binomial distribution, as described in “Advanced Topics in SPC” Wheeler, p255 ?  If not, and assuming you have an adequate sampling technique, and that data is from a single source, use an…[Read more]

Time based processes are usually strongly skewed and commonly bimodal.
Attempts to fit distributions is one of six sigma’s many fallacies.  There is no need to attempt to fit distributions to such data, in a process improvement context. 
Learn to understand what the data in the histogram is telling you.
You can find a useful exercise her…[Read more]

The other posts are on the right track but first you should check your data for the assumptions of p charts.  If these assumptions are not met, the increased sensitivity of p charts may be creating out of control points.
There is no need to remove points.  Your process appears to be unstable.
Read Wheeler “Understanding SPC”.

It is refreshing to see you are plotting manually.  You will learn more this way rather than blindly using software like most people.  Software can very much lead people astray, such as the idiotic normal distributions I see plotted over so many histograms. 
Data Points   Bins
50-100          6-10
100-250        7-12
>250              10-20
 

The Central Limit Theorem was of interest to Shewhart but he did not use it in formulating control charts.  This has been the source of much misunderstanding, particularly by authors such as Montgomery.
There is no need for the CLT in practical process improvement.

Excellent.  The only error you make is the transform.  These are not necessary and can make analysis more difficult.  I have a lovely example of a time based histogram for a help desk, bimodal and skewed right, which when tranformed becomes bimodal and skewed left.
You would enjoy reading “Normality and the Process Behaviour Chart” by Whe…[Read more]

Emily,
“Natural limits” !  Lovely to see a student of Wheeler in the morass of ignorance here.
Have a look at the papers here:
http://www.spcpress.com/spcink.htm

You may find this helpful:
http://www.spcpress.com/ink_pdfs/Good%20Data,%20Bad%20Data.pdf

Chebychev assures us that every distribution has a 3 sigma coverage of at least 88.9%. 
Standard XmR charts do far better than this for most situations. For skewness squared less than 3, and any kurtosis value, you get 98% cover !
For practical purposes, you can forget about normal distibutions, Box Cox and the rest.

Many software packages make it easy to use too many detection rules, which produce excessive false alarms.  Stick to the basic rule 1.
There is no need to recalculate limits, unless you have made a major change to the system.  Control limits are not absolutes – their aim is to gain insight into the process.

“The seven QC Tools, if used skilfully, will enable 95% of workplace problems to be solved” – Kaoru Ishikawa 1989.
He is right too !!!

Capability means consistently capable of meeting specification.  To do this a process must be in control.  It makes no difference what type of distribution the data has … and you will never exactly know it anyway.

Why ??????

KKG,
Your MBA studies might also benefit by a full investigation of six sigma by looking at the history of six sigma. The origins of six sigma’s numbers will help people understand how the error’s have been created. A brief summary follows:

Bill Smith, a Motorola engineer claims that for uncontrolled processes “batch to batch variation can b…[Read more]

Andy,
Shainin pre control isn’t “susceptible” to a shift – it allows for a movement in the mean in the green belt before its first level of correction procedures.  It doesn’t allow 1.5, it allows 50% of the spec limits –  1.5 happens to be the figure for Cp=1.0 (a rather uncommon situation), as we discussed previously.
Your suggestion of Bill…[Read more]

Well, it seems my post a week or so back was an accurate guess :
“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 ?  ”
A BA and a PhD in psychology !!!  No wonder there’s so much pseudo-statistical nonse…[Read more]