Pseudo-Scientific Versus DOE Approaches to Solving Problems
Our World Statistics Day conversations have been a great reminder of how much statistics can inform our lives. Do you have an example of how statistics has made a difference in your life? Share your story with the Community!
Experimenters may occasionally feel they lack sufficient time and/or resources to do a properly designed experiment, instead opting for a "Hail Mary" with a few experimental runs they intuitively feel may include the best solution. This presentation will walk the audience through a typical Hail Mary set of experimental runs that, at first glance, look promising. However, as we use Graph Builder to look at the experimental runs over time, look at scatterplots of the design space used, and use stepwise regression to attempt to build a model, we will realize that the data suggest a better set of factors might exist, but in a location that we didn’t originally test. We’ll evaluate the original design, learning that there is much confounding of potential interactions, and see very low optimality scores. We’ll then compare the original design to a simple DOE with no interactions, a response surface and a definitive screening design. We’ll conclude that the Hail Mary approach is lacking, and a properly designed experiment will yield better results using a similar amount of time and resources. The presentation will use JMP exclusively and will not rely on any PowerPoint slides.
The journal used during the presentation is attached.