Abstract

In today's fiercely competitive marketplace, successful and profitable companies distinguish themselves by bringing new products to market before their competitors. The cycle time to develop and launch new products largely depends on a company's ability to study large numbers of factors and to separate, or detect, the significant factors from the insignificant factors. Most ordinary experimental situations with many variables are easily satisfied with the use of a saturated, or nearly-saturated, fractional factorial experimental designs. However, there are occasions where the cost of mnning a statistically designed experiment can be so great as to prohibit the use of these techniques, forcing the experimenter to resort to other, riskier, experimental techniques. Theory suggests that a relatively new class of designs, systematic supersaturated designs, may prove to be even more effective at identifying significant factors than saturated, or nearly-saturated, fractional factorial designs. For the purpose of continuous improvement in the monetary and cycle time expenditures for new product design, new process launch, and new manufacturing process launch, supersaturated designs may provide the experimenter with a viable solution to the problem of studying more factors than permitted in a saturated design. Although, much has been written about creating supersaturated designs, little has been written regarding the analysis of these designs. This paper examines three test statistics which one might consider using when analyzing a supersaturated design. These test statistics are studied for four different supersaturated designs. The simulations and mathematical justifications presented in this paper suggest that it is not in the best interest of the experimenter to use these test statistics with these designs on a regular basis.

Library of Congress Subject Headings

Factorial experiment design; Principal components analysis; Factor analysis

Publication Date

8-1-1994

Document Type

Thesis

Department, Program, or Center

School of Mathematical Sciences (COS)

Advisor

Voelkel, Joseph

Comments

Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works. Physical copy available through RIT's The Wallace Library at: QA278.5 .K443 1994

Campus

RIT – Main Campus

Download

Share

COinS