Box-Cox transformations are used to transform a dataset that is not normally distributed. The transformed data is then fitted to a normal distribution and used to find a value for the capability of the process.

Nonnormal distributions in continuous data are typically associated with some form of boundary condition. Limits restrict the distribution in one direction. Good scenarios are where we have the boundary at 0. An example may include a measure of particle contamination in a packaging. The ultimate goal for medical devices would be to achieve zero particles; negative particle counts are not possible, and the closer we get to achieving the goal of zero, the more skewed our data can become.

Process times, again, are restricted to the 0 boundary; for example, we could study telephone answer times at a call centre. The time taken for an operator to pick up a call that is coming through to their phone cannot be negative; the phone must ring for...