Are You Leveraging Replicates and Repeats in Your DOEs?

Are You Leveraging Replicates and Repeats in Your DOEs?

Posted by Scott Sterbenz on Sep 18, 2018 8:42 am

We're always told to run replicates in our DOEs in order to increase the power of the DOE. But is that all the replicates can be used for? Absolutely not--they can be very powerful! Most times when we run a DOE, we want to optimize the response. I've run into the situation many times where there are several factor setting combinations that yield the optimum result, or at least values that are closely clustered to the absolute optimum. However, what you will find is that one set of factor combinations will yield a lot more variability in the response than another set of factor settings. So even though the mean of the response will be nearly identical, the variance may be significantly different. There are several tools we can use to study this. One is Taguchi's Signal to Noise ratio. Taguchi's philosophy was that it's better to be slightly off the optimum with less variation than it is to be right at the optimum with more variation. The first situation will yield much better results than the second.

At the upcoming Lean & Six Sigma Conference in Phoenix, I will be hosting a session on how to leverage Taguchi's Signal to Noise ratio in designed experiments. I show two industry examples of how Ford achieved great results using this technique. I'd love to see you there!
Best, Scott C. Sterbenz, P.E. ASQ Six Sigma Forum