Computational Statistics and Data Analysis 118 (2018) 84–97
Contents lists available at ScienceDirect
Computational Statistics and Data Analysis
journal homepage: www.elsevier.com/locate/csda
Optimizing two-level orthogonal arrays for simultaneously
estimating main effects and pre-specified two-factor
interactions
Ping-Yang Chen
a
, Ray-Bing Chen
a,
*, C. Devon Lin
b
a
Department of Statistics, National Cheng Kung University, Tainan 70101, Taiwan
b
Department of Mathematics and Statistics, Queen’s University, Kingston, ON, Canada
article info
Article history:
Received 16 November 2016
Received in revised form 7 July 2017
Accepted 9 August 2017
Available online 20 September 2017
Keywords:
D-optimal design
Fractional factorial design
Hadamard matrix
Swarm intelligence optimization
abstract
This paper considers the construction of D-optimal two-level orthogonal arrays that allow
for the joint estimation of all main effects and a specified set of two-factor interactions. A
sharper upper bound on the determinant of the related matrix is derived. To numerically
obtain D-optimal and nearly D-optimal orthogonal arrays of large run sizes, an efficient
search procedure is proposed based on a discrete optimization algorithm. Results on
designs of 20, 24, 28, 36, 44 and 52 runs with three or fewer two-factor interactions are
illustrated here to demonstrate the performance of the proposed approach. In addition,
two cases with four two-factor interactions are also demonstrated here.
© 2017 Elsevier B.V. All rights reserved.
1. Introduction
Two-level orthogonal arrays are most commonly used in industrial experiments to identify active effects because they
allow for the joint estimation of all main effects and interactions (Hamada and Wu, 1992). In some experiments, the
experimenter wants to estimate a set of specified effects, which is known as a requirement set (Greenfield, 1976) that consists
of all main effects and some specified two-factor interactions (2fis). For example, Wu and Chen (1992) provided an example
that the quality of products is affected by the factors in two independent manufacturing processes. In addition to the main
effects, the 2fis between factors from each of the two processes are necessary to take into account. For more details, please
refer Wu and Chen (1992).
There is ample research on obtaining two-level orthogonal arrays for estimating such a requirement set (see, for example,
Franklin and Bailey, 1977; Wu and Chen, 1992; Dey and Suen, 2002; Ke and Tang, 2003; Cheng and Tang, 2005). Recently,
Tang and Zhou (2009) investigated the existence of such orthogonal arrays and provided a construction strategy. Tang and
Zhou (2013) further derived theoretical results for finding D-optimal orthogonal arrays for jointly estimating main effects
and some specified 2fis, and showed the results of 12 and 20 runs for all the requirement sets with three or fewer 2fis.
The strategy for obtaining D-optimal orthogonal arrays in Tang and Zhou (2009, 2013) is as follows. Let S denote the
requirement set that includes m main effects and e specified 2fis, and C (S ) be a set that contains all 2fis in S and the main
effects that are involved in at least one 2fi in S . Thus C (S ) is a subset of S which is called the core set of S . For example,
for S ={F
1
, F
2
, F
3
, F
4
, F
5
, F
6
, F
1
F
2
, F
3
F
4
}, where F
i
F
j
is the 2fi of F
i
and F
j
, and then C (S ) ={F
1
, F
2
, F
3
, F
4
, F
1
F
2
, F
3
F
4
}. To find
an orthogonal array of n runs with m columns that supports S , a saturated orthogonal array with more columns, say H, is
considered, which can be expressed as (D
1
, D
2
, D
3
). Here the matrix D
1
is the subarray with m
1
columns that supports C (S ),
*
Corresponding author.
E-mail addresses: R28021017@mail.ncku.edu.tw (P.-Y. Chen), rbchen@mail.ncku.edu.tw (R.-B. Chen), cdlin@mast.queensu.ca (C.D. Lin).
http://dx.doi.org/10.1016/j.csda.2017.08.012
0167-9473/© 2017 Elsevier B.V. All rights reserved.