Abstract
The research of developing efficient methodologies for constructing optimal experimental designs has been very active in the last decade. Uniform design is one of the most popular approaches, carried out by filling up experimental points in a determinately uniform fashion. Applications of coding theory in experimental design are interesting and promising. Quaternary codes and their binary Gray map images attracted much attention from those researching design of experiments in recent years. The present paper aims at exploring new results for constructing uniform designs based on quaternary codes and their binary Gray map images. This paper studies the optimality of quaternary designs and their two and three binary Gray map image designs in terms of the uniformity criteria measured by: the Lee, wrap-around, symmetric, centered and mixture discrepancies. Strong relationships between quaternary designs and their two and three binary Gray map image designs are obtained, which can be used for efficiently constructing two-level designs from four-level designs and vice versa. The significance of this work is evaluated by comparing our results to the existing literature.
Similar content being viewed by others
References
Box GEP, Tyssedal J (1996) Projective properties of certain orthogonal arrays. Biometrika 83:950–955
Chatterjee K, Ou Z, Phoa FKH, Qin H (2017) Uniform four-level designs from two-level designs: a new look. Stat Sin 27:171–186
Chatterjee K, Li Z, Qin H (2012) Some new lower bounds to centered and wrap-round \(L_{2}\)-discrepancies. Stat Probab Lett 82:1367–1373
Deng LY, Tang B (1999) Generalized resolution and minimum aberration criteria for Plackett–Burman and other nonregular factorial designs. Stat Sin 9:1071–1082
Elsawah AM (2017a) A closer look at de-aliasing effects using an efficient foldover technique. Statistics 51(3):532–557
Elsawah AM (2017b) Constructing optimal router bit life sequential experimental designs: new results with a case study. Commun Stat Simul Comput. https://doi.org/10.1080/03610918.2017.1397164.
Elsawah AM, Qin H (2014) New lower bound for centered \(L_2\)-discrepancy of four-level \(U\)-type designs. Stat Probab Lett 93:65–71
Elsawah AM, Qin H (2015a) Lower bound of centered \(L_2\)-discrepancy for mixed two and three levels U-type designs. J Stat Plan Inference 161:1–11
Elsawah AM, Qin H (2015b) Mixture discrepancy on symmetric balanced designs. Stat Probab Lett 104:123–132
Elsawah AM, Qin H (2015c) A new strategy for optimal foldover two-level designs. Stat Probab Lett 103:116–126
Elsawah AM, Qin H (2016) Asymmetric uniform designs based on mixture discrepancy. J Appl Stat 43(12):2280–2294
Elsawah AM, Qin H (2017) Optimum mechanism for breaking the confounding effects of mixed-level designs. Comput Stat 32(2):781–802
Fang KT (1980) The uniform designs: application of number-theoretic methods in experimental design. Acta Math Appl Sin 3:363–372
Fang KT, Li RZ, Sudjianto A (2006a) Design and modeling for computer experiments. Chapman and Hall/CRC, New York
Fang KT, Maringer D, Tang Y, Winker P (2006b) Lower bounds and stochastic optimization algorithms for uniform designs with three or four levels. Math Comput 75:859–878
Fang KT, Tang Y, Yin J (2005) Lower bounds for the wrap-around \(L_2\)- discrepancy of symmetrical uniform designs. J Complex 21:757–771
Fang KT, Mukerjee R (2000) A connection between uniformity and aberration in regular fractions of two-level factorials. Biometrika 87:173–198
Fang KT, Lu X, Winker P (2003) Lower bounds for centered and wrap-around \(L_2\)-discrepancies and construction of uniform designs by threshold accepting. J Complex 19:692–711
Hickernell FJ (1998a) A generalized discrepancy and quadrature error bound. Math Comput 67:299–322
Hickernell FJ (1998b) Lattice rules: how well do they measure up? In: Hellekalek P, Larcher G (eds) Random and quasi-random point sets. Lecture Notes in Statistics, vol 138. Springer, New York, pp 109–166
Jiang B, Ai M (2014) Construction of uniform designs without replications. J Complex 30:98–110
Mukerjee R, Wu CFJ (2006) A modern theory of factorial designs. Springer, New York
Phoa FKH (2012) A code arithmetic approach for quaternary code designs and its application to (1/64)th-fraction. Ann Stat 40:3161–3175
Phoa FKH, Mukerjee R, Xu H (2012) One-eighth- and one-sixteenth-fraction quaternary code designs with high resolution. J Stat Plan Inference 142:1073–1080
Phoa FKH, Xu H (2009) Quarter-fraction factorial designs constructed via quaternary codes. Ann Stat 37:2561–2581
Qin H, Ai M (2007) A note on connection between uniformity and generalized minimum aberration. Stat Pap 48:491–502
Tang B, Deng LY (1999) Minimum \(G_2\)-aberration for nonregular fractional factorial designs. Ann Stat 27:1914–1926
Wang Y, Fang KT (1981) A note on uniform distribution and experimental design. Chin Sci Bull 26:485–489
Xu H, Phoa FKH, Wong WK (2009) Recent developments in nonregular fractional factorial designs. Stat Surv 3:18–46
Xu H, Wong A (2007) Two-level nonregular designs from quaternary linear codes. Stat Sin 17:1191–1213
Zhang RC, Phoa FKH, Mukerjee R, Xu H (2011) A trigonometric approach to quaternary code designs with application to one-eighth and one-sixteenth fractions. Ann Stat 39:931–955
Zhou YD, Ning JH, Song XB (2008) Lee discrepancy and its applications in experimental designs. Stat Probab Lett 78:1933–1942
Zhou YD, Fang KF, Ning JH (2013) Mixture discrepancy for quasi-random point sets. J Complex 29:283–301
Acknowledgements
The author greatly appreciate helpful suggestions of the two referees, the associate editor and the editor of chief Prof. Hajo Holzmann that significantly improved the paper. This work was partially supported by the UIC Grant (Nos. R201409 and R201712) and the Zhuhai Premier Discipline Grant.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
On behalf of all authors, the corresponding author states that there is no conflict of interest.
Appendix
Appendix
Proof of Theorem 1
The proof is obvious by the same technique as that in Elsawah and Qin (2015a, b). \(\square \)
Proof of Theorem 2
For any quaternary design \(\mathcal {U}_q\in \mathbb {U}(n; 4^m),\) let \(\beta _{ij},~\alpha _{ij},~\kappa _{ij},~\rho _{ij}, \ell _{ij}\) and \(\gamma _{ij}\) as given in Theorem 1. Then, we have
- (a):
-
\(\beta _{ij}+\alpha _{ij}+\ell _{ij}+\gamma _{ij}+\kappa _{ij}+\rho _{ij}=m,\)
- (b):
-
\(\sum _{i=1}^{n}\sum _{j\ne i}^{n}\beta _{ij}=\sum _{i=1}^{n}\sum _{j\ne i}^{n}\alpha _{ij}=\frac{1}{8}mn(n-4),\)
- (c):
-
\(\sum _{i=1}^{n}\beta _{ii}=\sum _{i=1}^{n}\alpha _{ii}=\frac{1}{2}mn,~\beta _{ii}+\alpha _{ii}=m,~\ell _{ii}=\gamma _{ii}=\kappa _{ii}=\rho _{ii}=0,\)
- (d):
-
\(\sum _{i=1}^{n}\sum _{j\ne i}^{n}\ell _{ij}=\sum _{i=1}^{n}\sum _{j\ne i}^{n}\gamma _{ij}=2\sum _{i=1}^{n}\sum _{j\ne i}^{n}\kappa _{ij}=2\sum _{i=1}^{n}\sum _{j\ne i}^{n}\rho _{ij}=\frac{1}{4}mn^2.\)
Thus, we can rewrite the analytical expressions of the uniformity criteria in Theorem 1 as follows
Define \(\zeta _{l}=1-\min \left\{ \frac{|x_{ik}-x_{jk}|}{4},1-\frac{|x_{ik}-x_{jk}|}{4}\right\} ,\) \(\zeta _{w}=\frac{3}{2}-|y_{ik}-y_{jk}|(1-|y_{ik}-y_{jk}|), ~\zeta _{c}=1+\frac{1}{2}|y_{ik}-\frac{1}{2}|+\frac{1}{2}|y_{jk}-\frac{1}{2}|-\frac{1}{2}|y_{ik}-y_{jk}|,~\zeta _{m}=\frac{15}{8}-\frac{1}{4}|y_{ik}-\frac{1}{2}|-\frac{1}{4}|y_{jk}-\frac{1}{2}|-\frac{3}{4}|y_{ik}-y_{jk}|+\frac{1}{2}|y_{ik}-y_{jk}|^2,~ \zeta ^*_{c}=1+\frac{1}{2}|y_{ik}-\frac{1}{2}|-\frac{1}{2}|y_{ik}-\frac{1}{2}|^{2}\) and \(\zeta ^*_{m}=\frac{5}{3}-\frac{1}{4}|y_{ik}-\frac{1}{2}|-\frac{1}{4}|y_{ik}-\frac{1}{2}|^{2}.\) Table 12 gives the distributions of \(\zeta _{l},~\zeta _{w},~\zeta _{c},~\zeta _{m},~\zeta ^*_{c}\) and \(\zeta ^*_{m},\) which are very useful to find the lower bounds of these discrepancies for quaternary designs. Thus, the expressions of the lower bounds of the uniformity criteria are straightforward according to the analytical expressions (7.1)–(7.4), the relations (a)–(d), Table 12 and based on the following two inequalities. Suppose \(\sum _{i=1}^{n}x_{i} = c,~\theta =n(\lfloor \frac{c}{n}\rfloor +1)-c,~x_{i}\)’s are nonnegative integers, then for any positive \(\mu ,\) we have \(\sum _{i=1}^{n}\mu ^{x_{i}}\ge \mu ^{\lfloor \frac{c}{n}\rfloor }(\theta +(n-\theta )\mu )\) (cf. Lemma 4 in Elsawah and Qin 2015c). Suppose \(\sum _{i=1}^{n}z_{i} = c_1,~\sum _{i=1}^{n}y_{i} = c_2,\) for \(i=1,2,\ldots ,n,\) where \(z_{i} ,~y_{i}\)’s are nonnegative integers and \(\alpha _s=n(\lfloor \frac{c_s}{n}\rfloor +1)-c_s,~s=1,2,\) then for any positive \(\mu _1\) and \(\mu _2,\) we have \(\sum _{i=1}^{n}\mu _1^{z_i}\mu _2^{y_i}\ge \left\{ \begin{array}{ll} (n-\alpha _1)e^{\chi _1}+(n-\alpha _2)e^{\chi _2}+(\alpha _1+\alpha _2-n)e^{\chi _3},~ \alpha _1+\alpha _1>n,\\ \alpha _2e^{\chi _1}+\alpha _1e^{\chi _2}+(n-\alpha _1-\alpha _2)e^{\chi _4},~ \alpha _1+\alpha _2\le n, \end{array} \right. \) where \(\chi _1=\varepsilon _1(\lfloor \frac{c_1}{n}\rfloor +1)+\varepsilon _2\lfloor \frac{c_2}{n}\rfloor ,~\chi _2=\varepsilon _1\lfloor \frac{c_1}{n}\rfloor +\varepsilon _2(\lfloor \frac{c_2}{n}\rfloor +1),~\chi _3=\varepsilon _1\lfloor \frac{c_1}{n}\rfloor +\varepsilon _2\lfloor \frac{c_2}{n}\rfloor ,~\chi _4=\varepsilon _1(\lfloor \frac{c_1}{n}\rfloor +1)+\varepsilon _2(\lfloor \frac{c_2}{n}\rfloor +1)\) and \((n-\alpha _1)\chi _1+(n-\alpha _2)\chi _2+(\alpha _1+\alpha _2-n)\chi _3=\alpha _2\chi _1+\alpha _1\chi _2+(n-\alpha _2-\alpha _1)\chi _4=\varepsilon _1c_1+\varepsilon _2c_2,~\varepsilon _1=\ln \mu _1\) and \(\varepsilon _2=\ln \mu _2\) (cf. Lemma 4 in Elsawah and Qin 2014). Utilizing the same technique as that in Elsawah and Qin (2015b) and Elsawah (2017a) with some algebra, we can complete the proof. \(\square \)
Proof of Theorem 3
The analytical expression \([{\mathcal {VD}^*}(\mathcal {U}_{2bi})]^2\) can be found in Theorem 1 in Elsawah and Qin (2015c). The proof of the analytical expression \([{\mathcal {VD}^{**}}(\mathcal {U}_{2bi})]^2\) is from Lemma 1 in Jiang and Ai (2014) and the discussion in Sect. 3 in Chatterjee et al. (2017) with some algebra, where we extend the existing results from two discrepancies, the \(\mathcal {CD}\) and \(\mathcal {WD},\) to all the five discrepancies, the \(\mathcal {LD},~\mathcal {WD},~\mathcal {SD},~\mathcal {CD}\) and \(\mathcal {MD},\) as well as we rewrite the analytical expressions of all these discrepancies as only one analytical expression \([{\mathcal {VD}^{**}}(\mathcal {U}_{2bi})]^2.\) Note that, for this case we have
However, the proof of the analytical expression \([{\mathcal {VD}^{***}}(\mathcal {U}_{2bi})]^2\) is given as follows. Under the two binary Gray map image principle, any pair of points \((x^q_{ik},x^q_{jk})\) in any quaternary design \(\mathcal {U}_{q}\in \mathbb {U}(n; 4^{m})\) will be replaced by the two pairs \((x^{2bi}_{ik_1},x^{2bi}_{jk_1})\) and \((x^{2bi}_{ik_2},x^{2bi}_{jk_2})\) in the two binary Gray map image design \(f_2(\mathcal {U}_{q})=\mathcal {U}_{2bi}\in \mathbb {U}(n; 2^{2m})\) as given in Table 13. From Table 13, we can show that
Then, we get
and
Thus, from the analytical expression \([{\mathcal {VD}^*}(\mathcal {U}_{2bi})]^2\) we get
\(\square \)
Proof of Theorem 4
For any two binary Gray map image design \(\mathcal {U}_{2bi}\in \mathbb {U}(n;2^{2m}),\) let the Hamming distance \(h_{ij}=\mathcal {N}\left\{ (i,j):(x_{ik},x_{jk})\in \left\{ \left( \frac{1}{4},\frac{3}{4}\right) ,\left( \frac{3}{4},\frac{1}{4}\right) \right\} \right\} .\) Then, we have
The lower bound \({\mathcal {LBV}^*_{2bi}}\) is from the analytical expression \([{\mathcal {VD}^{*}}(\mathcal {U}_{2bi})]^2\) in Theorem 3, the relation (7.5) and Lemma 4 in Elsawah and Qin (2015c). The proof of the lower bound \({\mathcal {LBV}^{**}_{2bi}}\) is from the new analytical expression \([{\mathcal {VD}^{**}}(\mathcal {U}_{2bi})]^2\) in Theorem 3, the proof of Lemma 3 in Chatterjee et al. (2017) and Lemma 2 in Chatterjee et al. (2012) with some algebra, where we extend the existing results from two discrepancies, the \(\mathcal {CD}\) and \(\mathcal {WD},\) to all the five discrepancies, the \(\mathcal {LD},~\mathcal {WD},~\mathcal {SD},~\mathcal {CD}\) and \(\mathcal {MD},\) as well as we express the lower bounds of all these discrepancies as only one analytical expression \([{\mathcal {LBV}^{**}}(\mathcal {U}_{2bi})]^2.\) However, the proof of the lower bound \({\mathcal {LBV}^{***}_{2bi}}\) is given as follows. From the relations (a) and (c) in the proof of Theorem 2, we can rewrite the analytical formulas of different discrepancies in \([{\mathcal {VD}^{***}}(\mathcal {U}_{2bi})]^2\) as follows
From the relations (b) and (d) in the proof of Theorem 2, we get
The proof now is obvious from (7.6) and (7.7) by using Lemma 4 in Elsawah and Qin (2014). \(\square \)
Proof of Theorem 5
The analytical expression \([{\mathcal {VD}^*}(\mathcal {U}_{3bi})]^2\) can be found in Theorem 1 in Elsawah and Qin (2015c). The proof of the analytical expression \([{\mathcal {VD}^{**}}(\mathcal {U}_{3bi})]^2\) is from Lemma 1 in Jiang and Ai (2014) and the discussion in Sect. 4 in Chatterjee et al. (2017) with some algebra, where we extend the existing results from two discrepancies, the \(\mathcal {CD}\) and \(\mathcal {WD},\) to all the five discrepancies, the \(\mathcal {LD},~\mathcal {WD},~\mathcal {SD},~\mathcal {CD}\) and \(\mathcal {MD},\) as well as we rewrite the analytical expressions of all these discrepancies as only one analytical expression \([{\mathcal {VD}^{**}}(\mathcal {U}_{3bi})]^2.\) Note that, for this case we have
However, the proof of the analytical expression \([{\mathcal {VD}^{***}}(\mathcal {U}_{3bi})]^2\) is given as follows. Under the three binary Gray map image principle, any pair of points \((x^q_{ik},x^q_{jk})\) in any quaternary design \(\mathcal {U}_{q}\in \mathbb {U}(n; 4^{m})\) will be replaced by the three pairs \((x^{3bi}_{ik_1},x^{3bi}_{jk_1})\), \((x^{3bi}_{ik_2},x^{3bi}_{jk_2})\) and \((x^{3bi}_{ik_3},x^{3bi}_{jk_3})\) in the three binary Gray map image design \(f_3(\mathcal {U}_{q})=\mathcal {U}_{3bi}\in \mathbb {U}(n; 2^{3m})\) as given in Table 14. From Table 14, we can show that
Then, we get
and
Thus, from the analytical expression \([{\mathcal {VD}^*}(\mathcal {U}_{3bi})]^2\) we get
\(\square \)
Proof of Theorem 6
For any three binary Gray map image design \(\mathcal {U}_{3bi}\in \mathbb {U}(n;2^{3m}),\) let the Hamming distance \(h_{ij}=\mathcal {N}\left\{ (i,j):(x_{ik},x_{jk})\in \left\{ \left( \frac{1}{4},\frac{3}{4}\right) ,\left( \frac{3}{4},\frac{1}{4}\right) \right\} \right\} .\) Then, we have
The lower bound \({\mathcal {LBV}^*_{3bi}}\) is from the analytical expression \([{\mathcal {VD}^{*}}(\mathcal {U}_{3bi})]^2\) in Theorem 5, the relation (7.8) and Lemma 4 in Elsawah and Qin (2015c). The proof of the lower bound \({\mathcal {LBV}^{**}_{3bi}}\) is from the new analytical expression \([{\mathcal {VD}^{**}}(\mathcal {U}_{3bi})]^2\) in Theorem 5, the proof of Lemma 5 in Chatterjee et al. (2017) and Lemma 2 in Chatterjee et al. (2012) with some algebra, where we extend the existing results from two discrepancies, the \(\mathcal {CD}\) and \(\mathcal {WD},\) to all the five discrepancies, the \(\mathcal {LD},~\mathcal {WD},~\mathcal {SD},~\mathcal {CD}\) and \(\mathcal {MD},\) as well as we express the lower bounds of all these discrepancies as only one analytical expression \([{\mathcal {LBV}^{**}}(\mathcal {U}_{3bi})]^2.\) However, the proof of the lower bound \({\mathcal {LBV}^{***}_{2bi}}\) is given as follows. From the relations (a) and (c) in the proof of Theorem 2, we can rewrite the analytical formulas of different discrepancies in \([{\mathcal {VD}^{***}}(\mathcal {U}_{3bi})]^2\) as follows
The proof now is obvious from (7.9) and (7.7) by using Lemma 4 in Elsawah and Qin (2015c). \(\square \)
Rights and permissions
About this article
Cite this article
Elsawah, A.M., Fang, KT. New results on quaternary codes and their Gray map images for constructing uniform designs. Metrika 81, 307–336 (2018). https://doi.org/10.1007/s00184-018-0644-5
Received:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00184-018-0644-5