|
Activity Number:
|
331
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 11, 2004 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Quality and Productivity
|
| Abstract - #301409 |
|
Title:
|
Comparing the Performance of Crossed-array and Single-array Approaches to Robust Design
|
|
Author(s):
|
Daniel D. Frey*+ and Xiang Li
|
|
Companies:
|
Massachusetts Institute of Technology and Massachusetts Institute of Technology
|
|
Address:
|
77 Massachusetts Ave. Room 3-449D, Cambridge, MA, 02139,
|
|
Keywords:
|
robust design ; design of experiments
|
|
Abstract:
|
This paper presents a new way of comparing the effectiveness of robust design methods and applies it to two well-known alternatives--crossed arrays and single arrays. The relative merits of both approaches have been examined in the literature and single arrays are generally considered the preferred alternative. A large number of simulated systems are sampled from a relaxed weak heredity model so that the systems will exibit the properties of effect sparsity, hierarchy, and inheritance. Crossed-array and single-array methods are applied to the systems in an effort to reduce the variance of the response. The percentage of variance reduction achieved on average across the sampled systems is taken as a measure of performance of the method. The simulations indicate that the single array approach performs well if the probability of three-way interactions is zero, but degrades if there is a realistic likelihood of control by control by noise interactions. By contrast, crossed-array approaches performed well across a broader range of system types.
|
