Jensen W.A.,W.L. Gore and Inc.
Journal of Quality Technology | Year: 2016
Throughout the literature on design of experiments, the need to perform confirmation runs has been repeatedly emphasized. We provide a high-level framework for design of experiments and show how the confirmation phase is becoming increasingly important. However, the literature provides little guidance on how to assess the confirmation runs nor how many confirmation runs to perform. We provide a review of the literature that addresses confirmation runs. We propose several methods for assessing confirmation runs based on statistical intervals. A simulation study shows which are the best approaches for assessing confirmation runs. We then give some guidance on how many confirmation runs are appropriate and what to do when one fails a confirmation experiment.