For many of you, this is a basic reminder, but it's an incredibly important one. Before jumping into any DOE or regression study--or even most hypothesis tests--ask the question: Is your process stable? DMAIC is meant for solving common cause variation. If your process is not stable and in control, you have more than just common cause variation, you probably have special cause variation! Starting any type of DOE or regression analysis at this point may result in bad decisions. Why? Special cause variation often reveals itself in the form of outliers. Imagine running a DOE or regression study, and during your data collection, you get one of these outliers. What is that going to do to your results? How is that going to affect the significance of your factors? A factor might come forth as significant when it is really not. Be sure to address that special cause variation before looking at running a DOE or regression study. You can check your data with a simple histogram and normality test, a control chart, or even a chi-square test for attribute data.
Scott C. Sterbenz, P.E.
ASQ Six Sigma Forum