|
Activity Number:
|
36
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 2, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #304216 |
|
Title:
|
A Computational Algorithm for Searching Optimal Saturated Two-Level Factorial Designs
|
|
Author(s):
|
Samad A. Hedayat and Haiyuan Zhu*+
|
|
Companies:
|
University of Illinois at Chicago and Merck Research Laboratories
|
|
Address:
|
128 E. Lincoln Av., Rahway, NJ, 08816,
|
|
Keywords:
|
Algorithms ; Factorial designs ; D-optimality ; D-Efficiency
|
|
Abstract:
|
Two level factorial designs are among the most often used designs in experiments. Saturated optimal two level designs are available for many cases, but for the cases that the optimal or highly efficient designs are unknown, computational algorithms are needed to search for them. In searching D-optimal or D-efficient designs, available software packages often need to calculate the determinant of the information matrices which demand massive computational times, and often fail for designs involving large number of factors. This paper introduces a new algorithm for searching saturated D-optimal and D-efficient two-level factorial designs. The new algorithm has a higher computational efficiency in comparison with other algorithms for this specific setup since it utilizes the geometrical formulation of the design matrices and thus bypassing the direct computation of the determinants.
|
