Split Plots: Modeling the Process Mean and Variance from a CCD

• Activity Number: 211
• Type: Topic Contributed
• Date/Time: Tuesday, August 10, 2004 : 10:30 AM to 12:20 PM
• Sponsor: Section on Quality and Productivity
• Author(s): Scott Kowalski*+ and Geoff Vining and Douglas Montgomery
• Companies: Minitab Inc. and Virginia Polytechnic Institute and State University and Arizona State University
• Keywords: split plot ; response surface methodology

Abstract

The response surface methodology (RSM) framework for designed experiments has been widely adopted in practice. Most of the literature on RSM is confined to completely randomized experimental designs (CRDs). Unfortunately, industrial experimenters often encounter situations where some experimental factors are hard to change or where there is significant discrepancy in the size of some experimental units. For both of these situations, the experimental units for some factors are observational units for other factors, which leads to split plot designs. An important question within industrial statistics is how to find operating conditions that achieve some goal for the mean of a characteristic of interest while simultaneously minimizing the characteristics process variance. This talk first establishes how one can modify the common central composite design to accommodate efficiently a split-plot structure. It then modifies this design to allow the estimation of separate models for the characteristic's mean and variance. The appropriate analysis of the experimental results will be discussed.