The distribution that shows up when I carry out a Capability Study is not Normal. Does this mean I can’t use conventional Control Charts and do I have to calculate Cp and Cpk in a different way?
A / There is much rubbish talked about this. Basically, if your characteristic is two sided (i.e. nominal + / – the same tolerance), then your Capability Study should present a Normal distribution, or at least a symmetrical distribution. If it doesn’t then something is peculiar. I would investigate before going any further. If on the other hand the characteristic is one-sided (commonly where the nominal and one of the limits is the same and / or there isn’t a second limit) then you should expect a skewed distribution. When you come to sampling for control charts, as long as your sample size is 5 or more, then the Central Limit Theorem comes into play viz, the distribution of the sample means (i.e. the shape of the control chart) will be Normal as long as you remain in control. So it doesn’t really matter what the original distribution is. As long as you are sampling at the rate of 5 don’t worry, conventional Xbar Range charts will work fine. Even a sample size of 3 gives a reasonable a
Related Questions
- The distribution that shows up when I carry out a Capability Study is not Normal. Does this mean I can’t use conventional Control Charts and do I have to calculate Cp and Cpk in a different way?
- The frozen bottle shows uneven color distribution and upon thawing it has layered color distribution. Is it normal?
- What does normal or transformed-normal mean when referring to the distribution of the data?
