Talk:SPC: Basic control charts: theory and construction, sample size, x-bar, r charts, s charts


 * general question, where do the parameters A2 and the like come from? Can they be derived?
 * Can you use a sampling approach to create data and show that it is in or out of control?

--MthompsoNZ 01:14, 18 July 2011 (EDT)I would like to offer a different presentation of the table of constants. Having been trained by Ford Motor Company, we were drilled in the fact that there is no range lower control line at 0, but that there is 'none' there (until the sample size is large enough). Hence, I would prefer the table to read 'None' for the B3 and D3 values that are currently showing 0. If I can dig up my old Ford books I will give you a reference, but it's so long since I've practiced that they may even be given away by now. However, in my defence, I can say that the Ford approach to SPC was so good it was an agreed approach for the big three auto manufacturers, and was adopted by Boeing.

To the questions previously asked, here is my answer: The control lines calculated using A2 are basically the 3-sigma limits on the distribution of the sample means. As such, sigma(means) = sigma(population)/square root of sample size (how do I insert a formula - I click the button, then what?) Also, the control chart approach depends on having each sample from an 'instant' of time, as far as possible. You cannot take a batch, sample from that from random time slots, and then see if it was 'in control'. You can, of course, sample and then assess the distribution, but the control chart is much more powerful than that.

Oh yes, and before you take my word too seriously, I am what Dr Deming calls 'the bane of the profession' - a practical statistician. I have taught SPC to illiterate operators, who have gone out smiling at what worlds open up once they have 'control' of their processes.