Variation in Controlled Experimental Variables

Abstract

When a designed experiment is conducted, the exact levels of controlled experimental variables specified by the design may not be attainable. Two different types of situations can occur in which the actual settings of the design variables deviate from the target design settings. In one case, the actual settings obtained are unknown. In the other case, the actual settings are known; however desirable design properties such as orthogonality may have been lost. This paper examines the impact of known and unknown errors in experimental variables for designed experiments in which the response is assumed to be approximately linear in the region of interest. For comparison, the well-known situation where the target values of the experimental variables can actually be obtained will also be considered.

Key Words:

• Experiment design
• efficiency
• errors
• orthogonality
• uncertainty.

Additional information

Notes on contributors

Joanne R. Wendelberger

Joanne Wendelberger is the Group Leader of the Statistical Sciences Group at Los Alamos National Laboratory. She received her Ph.D. in Statistics in 1991 from the Department of Statistics, University of Wisconsin-Madison, United States, under the advisement of Professor Geoge Box. Her current research interests include statistical experimental design and test planning, statistical bounding and uncertainty quantification, sampling and analysis for large-scale computation and visualization, and probabilistic computing. Dr. Wendelberger is a Fellow of the American Statistical Association.