Common experimental practices suggest randomizing the order in which runs are performed. However, there may be situations in which randomization might not produce the most desirable order, especially in the presence of known trends. There has been research done on systematically designing experiments to be robust against trends. However, few studies address the additional dimensions that arise in nested designs such as split-plot designs. Split-plot designs have been used for many years in agricultural applications and are sometimes preferred where there are hard-to-change factors in industrial settings. There currently is no established methodology to produce split-plot designs that are robust to potential two-dimensional trends. The objective of this work is to develop a methodology to design run orders for two-level, split-plot (2w × 2s) designs that are robust or nearly robust against a set of trends. Two methods are developed in this work. A fold-over method that uses already established principles is extended for use in split-plot designs. The second method uses an integer linear programming approach to search for an optimal design that is resistant to specific trends. A comparison between the two methods is presented and evaluated with a proposed set of metrics.

Library of Congress Subject Headings

Experimental design; Linear programming; Integer programming

Publication Date


Document Type



Carrano, Andres

Advisor/Committee Member

Thorn, Brian


Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works in December 2013. Physical copy available through RIT's The Wallace Library at: QA279 .L67 2007


RIT – Main Campus