Abstract
In many practical applications, follow-up experimental designs are commonly used to explore the relationship between inputs and outputs steps by steps. As some additional resources could be available to the experimenter after the first step, some additional runs or factors may be added in the follow-up stage. The issue of uniform augmented q-level designs is investigated in this paper. Using the level permutation technique, the uniformity of augmented q-level designs is discussed under the average mixture discrepancy, and new lower bounds of average mixture discrepancy for augmented designs are obtained, which can be served as the benchmark for searching uniform augmented designs. Numerical examples show that the uniform augmented designs can be constructed with high efficiency.
Similar content being viewed by others
References
Chen W, Qi ZF, Zhou YD (2015) Constructing uniformity designs under mixture discrepancy. Stat Prob Lett 97:76–82
Fang KT, Wang Y (1994) Number-theoretic methods in statistics. Chapman and Hall, London
Fang KT, Lu X, Winker P (2003) Lower bounds for centered and wrap-around \(L_2\)-discrepancy and construction of uniform designs by threshold accepting. J Complex 19:692–711
Fang KT, Li R, Sudjianto A (2005) Design and modeling for computer experiments. Chapman and Hall, London
Fang KT, Liu MQ, Qin H, Zhou YD (2018) Theory and application of uniform experimental designs, vol 221. Lecture notes in statistics. Spinger, Singapore
Gou TX, Qin H, Chatterjee K (2017) A new extension strategy on three-level factorials under wrap-around \(L_2\)-discrepancy. Commun Stat Theory Method 46:9063–9074
Gupta VK, Singh P, Kole B, Parsad R (2010) Addition of runs to a two-level supersaturated design. J Stat Plann Inference 140:2531–2535
Gupta VK, Chatterjee K, Das A, Kole B (2012) Addition of runs to an \(s\)-level supersaturated design. J Stat Plann Inference 142:2402–2408
Hu LP, Chatterjee K, Liu JQ, Ou ZJ (2020) New Lower bound for Lee discrepancy of asymmetrical factorials. Stat Pap 61:1763–1772
Liu JQ, Ou ZJ, Hu LP, Wang K (2019a) Lee discrepancy on mixed two-and three-level uniform augmented designs. Commun Stat Theory Methods 48(10):2409–2424
Liu JQ, Ou ZJ, Li HY (2019b) Uniform row augmented designs with multi-level. Commun Stat Theory Methods. https://doi.org/10.1080/03610926.2019.1705979
Ou ZJ, Qin H (2019) Optimal foldover plans of asymmetric factorials with minimum wrap-around \(L_2\)-discrepancy. Stat Pap 60(5):1699–1716
Ou ZJ, Qin H, Li HY (2011) Optimal blocking and foldover plans for non-regular two-level designs. J Stat Plann Inference 141(5):1635–1645
Ou ZJ, Qin H, Cai X (2013) Partially replicated two-level fractional factorial designs via semifoldover. J Stat Plann Inference 143(4):809–817
Qin H, Chatterjee K, Ghosh S (2015) Extended mixed-level supersaturated designs. J Stat Plann Inference 157–158:100–107
Qin H, Gou T, Chatterjee K (2016) A new class of two-level optimal extended designs. J Korean Stat Soc 45:168–175
Tang Y, Xu HQ (2013) An effective construction method for multi-level uniform designs. J Stat Plann Inference 143(9):1583–1589
Tang Y, Xu HQ, Lin DKJ (2012) Uniform fractional factorial designs. Ann Stat 40:891–907
Wang Y, Fang KT (1981) A note on uniform distribution and experimental design. Chin Sci Bull 26(6):485–489
Wang K, Ou ZJ, Liu JQ, Li HY (2020) Uniformity pattern of \(q\)-level factorials under mixture discrepancy. Stat Pap. https://doi.org/10.1007/s00362-019-01155-2
Yang F, Zhou YD, Zhang XR (2017) Augmented uniform designs. J Stat Plann Inference 182:168–175
Yang F, Zhou YD, Zhang AJ (2019) Mixed-level column augmented uniform designs. J Complex 53:23–39
Zhang QH, Wang ZH, Hu JW, Qin H (2015) A new lower bound for wrap-around \({L_2}\)-discrepancy on two and three mixed level factorials. Stat Probab Lett 96:133–140
Zhou YD, Fang KT (2013) An efficient method for constructing uniform designs with large size. Comput Stat 28:1319–1331
Zhou YD, Xu HQ (2014) Space-filling fractional designs. J Am Stat Assoc 109(57):1134–1144
Zhou YD, Fang KT, Ning JH (2013) Mixture discrepancy for quasi-random point sets. J Complex 29(3–4):283–301
Acknowledgements
The authors greatly appreciate helpful suggestions of an associate editor and the referees. This work was partially supported by the National Natural Science Foundation of China (Grant Nos. 11701213, 11961027, 11561025), Hunan Provincial Natural Science Foundation of China (Grant Nos. 2020JJ4497, 2019JJ40041), Scientific Research Plan Item of Hunan Provincial Department of Education (Grant Nos. 18A284, 19A403).
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
Proof of Proposition 1
The proof of Proposition 1 is similar to the Lemma 1 of both Zhou and Xu (2014) and Wang et al. (2020), based on the notations of \(\varphi _{ij}\), one can combine the formula (1) with formula (2) and demonstrate that
and
where
Then, the formula (2) can be rewritten as
-
(i)
for odd q,
$$\begin{aligned}&\left[ \overline{MD}({{\varvec{d}}})\right] ^2\nonumber \\&\quad = \frac{1}{(q!)^s}\sum _{{{\varvec{d}}}'\in {{\mathcal {P}}}({{\varvec{d}}})} \left[ \left( \frac{7}{12}\right) ^{s}-\frac{2}{n}\sum _{i=1}^n\prod _{k=1}^s f_1(x_{ik})+\frac{1}{n^2}\sum _{i=1}^n\sum _{j=1}^n\prod _{k=1}^sf(x_{ik},x_{jk})\right] \nonumber \\&\quad = \left( \frac{7}{12}\right) ^s-2\left( \frac{7q^2+1}{12q^2}\right) ^s +\frac{1}{n^2}\left( \frac{14q^2-4q+3}{24q^2}\right) ^s \sum _{i=1}^n\sum _{j=1}^n \alpha ^{\varphi _{ij}}\nonumber \\&\quad = \left( \frac{7}{12}\right) ^s-2\left( \frac{7q^2+1}{12q^2}\right) ^s +\frac{1}{n^2}\left( \frac{14q^2-4q+3}{24q^2}\right) ^s\nonumber \\&\quad \qquad \left[ \sum _{i=1}^n\sum _{j(=i)=1}^n\alpha ^{\varphi _{ij}} +\sum _{i=1}^n\sum _{j(\ne i)=1}^n\alpha ^{\varphi _{ij}}\right] \nonumber \\&\quad = \left( \frac{7}{12}\right) ^s-2\left( \frac{7q^2+1}{12q^2}\right) ^s +\frac{1}{n}\left( \frac{6q^2+1}{8q^2}\right) ^s\nonumber \\&\qquad +\frac{1}{n^2}\left( \frac{14q^2-4q+3}{24q^2}\right) ^s \sum _{i=1}^n\sum _{j(\ne i)=1}^n \alpha ^{\varphi _{ij}}, \end{aligned}$$(18) -
(ii)
similarly, for even q,
$$\begin{aligned}&\left[ \overline{MD}({{\varvec{d}}})\right] ^2\nonumber \\&\quad = \left( \frac{7}{12}\right) ^s -2\left( \frac{28q^2+1}{48q^2}\right) ^s +\frac{1}{n^2}\left( \frac{7q-2}{12q}\right) ^s \sum _{i=1}^n\sum _{j=1}^n \alpha ^{\varphi _{ij}}\nonumber \\&\quad = \left( \frac{7}{12}\right) ^s -2\left( \frac{28q^2+1}{48q^2}\right) ^s\nonumber \\&\qquad +\frac{1}{n^2}\left( \frac{7q-2}{12q}\right) ^s \left[ \sum _{i=1}^n\sum _{j(=i)=1}^n\alpha ^{\varphi _{ij}} +\sum _{i=1}^n\sum _{j(\ne i)=1}^n\alpha ^{\varphi _{ij}}\right] \nonumber \\&\quad = \left( \frac{7}{12}\right) ^s -2\left( \frac{28q^2+1}{48q^2}\right) ^s +\frac{1}{n}\left( \frac{3}{4}\right) ^s+\frac{1}{n^2}\left( \frac{7q-2}{12q}\right) ^s \sum _{i=1}^n\sum _{j(\ne i)=1}^n \alpha ^{\varphi _{ij}},\nonumber \\ \end{aligned}$$(19)which completes the proof. \(\square \)
Proof of Theorem 1
Based on Proposition 1 and Lemma 1, as for the portion of the fourth term in (3) and (4), one can obtain
which completes the proof. \(\square \)
Proof of Theorem 2
According to Theorem 1 and Proposition 2, based on the notations of \(\varphi _{ij}\), similar to (20), we can obtain that
Based on the Eqs. (9) and (10) in Proposition 2 and inequalities (21)-(23), Theorem 2 is obviously, which completes the proof. \(\square \)
Proof of Proposition 3
Based on the formula (14), for odd q, we are able to easily get
According to the formula (18), we have
Thus, the formula (15) can be easily obtained by substituting Eq. (25) into the formula (24). Similarly, the formula (16) can be obtained as follow, while q is even
which completes the proof. \(\square \)
Rights and permissions
About this article
Cite this article
Hu, Z., Liu, J., Li, Y. et al. Uniform augmented q-level designs. Metrika 84, 969–995 (2021). https://doi.org/10.1007/s00184-020-00793-z
Received:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00184-020-00793-z