|
Activity Number:
|
214
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
| Abstract - #305595 |
|
Title:
|
Comparison of Screening Methods in the Presence of Interactions
|
|
Author(s):
|
Danel Draguljic*+ and Dave Woods and Angela Dean and Susan Lewis
|
|
Companies:
|
The Ohio State University and University of Southampton and The Ohio State University and University of Southampton
|
|
Address:
|
1958 Neil Avenue, Columbus, OH, 43210-1247,
|
|
Keywords:
|
supersaturated design ; group screening
|
|
Abstract:
|
In the presence of a large number of factors, fractional factorial designs may be infeasible due to the need of estimation of a large number of effects. An approach is to perform a "screening experiment" followed by a conventional experiment on a subset of the factors. Ideally, a screening experiment sifts through all effects of interest and identifies the few that influence the response substantially. An effective screening tool is a supersaturated design; a design which contains fewer runs than number of effects to be estimated. This discussion will briefly describe cyclic methods of construction of supersaturated designs and present and compare the analysis methods proposed in the literature. Also, a two-stage group screening method will be included in the comparison. The comparison tests the ability of methods to identify correctly the significant effects under a range of scenarios.
|
