In the previous recipes, we used normal distribution to estimate capability. When data is not distributed normally, the normal distribution will incorrectly estimate the amount of the results that we find outside of the specifications, and it makes the calculation of the capability inaccurate. Here, we will look to find an appropriate nonnormal distribution or transformation to fit to the data.
It is vital to indicate that before looking at transformations or nonnormal distributions, we must understand the process that we are investigating. We must consider the reasons for the data not being normal. An unstable process can fail the normality test because the mean or variation is moving over time. There may be trends or process shifts. With this data, we should investigate the issues of the unstable process before using transformations as a fix.
Other issues may be that we have several distributions within the block of data. Different machines...