|
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 - #303663 |
|
Title:
|
24-Run Hadamard Designs for Estimation and Detection of Interactions
|
|
Author(s):
|
Yingfu F. Li*+
|
|
Companies:
|
University of Houston-Clear Lake
|
|
Address:
|
2700 Bay Area Blvd, Houston, TX, 77058,
|
|
Keywords:
|
Estimation ; Hadamard design ; two-factor interactions
|
|
Abstract:
|
The folded-over 12-run Plackett-Burman design ($PB_{12+12}$) has been intensively studied in Diamond (1995) and Miller and Sitter (2001). They concluded that such a design is well suited for detecting significant two-factor interactions in some screening applications. $PB_{12+12}$ can also be constructed from Hadamard matrices (Hadamard design) of order 24. However, we observed that many such 24-run Hadamard designs have much better estimability than that of $PB_{12+12}$. In this paper, we thoroughly investigate all possible designs from 60 non-equivalent Hadamard matrices of order 24 in the sense of estimation and detection of the maximum possible number of two-factor interactions in addition to all main effects. The simulation study shows in certain situations some our obtained 24-run Hadamard designs are better choices over the $PB_{12+12}$.
|
