How to Plan and Analyze a Verification DOE


presented by Shari Kraber

Companies are gradually learning that design of experiments (DOE) can be a useful tool during the final verification stage of a product or process. Rather than just testing the extremes one factor at a time, a DOE can cover the expected range of production variation and confirm that the final production conditions will make product within specification. The underlying difficulty in this application is that the desired DOE result is to have NO significant effects. This talk demonstrates how to confirm that the DOE chosen has the correct power to detect effects IF they exist. The design must be capable of doing the job. After collecting data a variety of potential results to estimate power are presented on half-normal plots. The interpretation of each plot will be discussed.