Skip to main content
Log in

Uniform augmented q-level designs

  • Published:
Metrika Aims and scope Submit manuscript

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.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

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

    Article  Google Scholar 

  • Fang KT, Wang Y (1994) Number-theoretic methods in statistics. Chapman and Hall, London

    Book  Google Scholar 

  • 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

    Article  Google Scholar 

  • Fang KT, Li R, Sudjianto A (2005) Design and modeling for computer experiments. Chapman and Hall, London

    Book  Google Scholar 

  • Fang KT, Liu MQ, Qin H, Zhou YD (2018) Theory and application of uniform experimental designs, vol 221. Lecture notes in statistics. Spinger, Singapore

    Book  Google Scholar 

  • 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

    Article  Google Scholar 

  • 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

    Article  MathSciNet  Google Scholar 

  • 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

    Article  MathSciNet  Google Scholar 

  • Hu LP, Chatterjee K, Liu JQ, Ou ZJ (2020) New Lower bound for Lee discrepancy of asymmetrical factorials. Stat Pap 61:1763–1772

    Article  MathSciNet  Google Scholar 

  • 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

    Article  Google Scholar 

  • 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

    Article  Google Scholar 

  • Ou ZJ, Qin H, Cai X (2013) Partially replicated two-level fractional factorial designs via semifoldover. J Stat Plann Inference 143(4):809–817

    Article  MathSciNet  Google Scholar 

  • Qin H, Chatterjee K, Ghosh S (2015) Extended mixed-level supersaturated designs. J Stat Plann Inference 157–158:100–107

    Article  MathSciNet  Google Scholar 

  • Qin H, Gou T, Chatterjee K (2016) A new class of two-level optimal extended designs. J Korean Stat Soc 45:168–175

    Article  MathSciNet  Google Scholar 

  • Tang Y, Xu HQ (2013) An effective construction method for multi-level uniform designs. J Stat Plann Inference 143(9):1583–1589

    Article  MathSciNet  Google Scholar 

  • Tang Y, Xu HQ, Lin DKJ (2012) Uniform fractional factorial designs. Ann Stat 40:891–907

    MathSciNet  MATH  Google Scholar 

  • Wang Y, Fang KT (1981) A note on uniform distribution and experimental design. Chin Sci Bull 26(6):485–489

    MathSciNet  MATH  Google Scholar 

  • 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

    Article  Google Scholar 

  • Yang F, Zhou YD, Zhang XR (2017) Augmented uniform designs. J Stat Plann Inference 182:168–175

    Article  MathSciNet  Google Scholar 

  • Yang F, Zhou YD, Zhang AJ (2019) Mixed-level column augmented uniform designs. J Complex 53:23–39

    Article  MathSciNet  Google Scholar 

  • 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

    Article  Google Scholar 

  • Zhou YD, Fang KT (2013) An efficient method for constructing uniform designs with large size. Comput Stat 28:1319–1331

    Article  MathSciNet  Google Scholar 

  • Zhou YD, Xu HQ (2014) Space-filling fractional designs. J Am Stat Assoc 109(57):1134–1144

    Article  MathSciNet  Google Scholar 

  • Zhou YD, Fang KT, Ning JH (2013) Mixture discrepancy for quasi-random point sets. J Complex 29(3–4):283–301

    Article  MathSciNet  Google Scholar 

Download references

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

Authors

Corresponding author

Correspondence to Hongyi Li.

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

$$\begin{aligned} \sum _{{{\varvec{d}}}'\in {{\mathcal {P}}}({{\varvec{d}}})}\sum _{i=1}^n\prod _{k=1}^sf_1(x_{ik})= & {} n\left[ (q-1)!{\sum _{j=0}^{q-1}}f_1\left( \frac{2j+1}{2q}\right) \right] ^{s} \end{aligned}$$

and

$$\begin{aligned} \sum _{{{\varvec{d}}}'\in {{\mathcal {P}}}({{\varvec{d}}})}\sum _{i=1}^n\sum _{j=1}^n\prod _{k=1}^sf(x_{ik},x_{jk})= & {} (q!)^s\sum _{i=1}^n\sum _{j=1}^n\left[ \frac{\alpha \beta }{q(q-1)}\right] ^{\varphi _{ij}} \left[ \frac{\beta }{q(q-1)}\right] ^{s-\varphi _{ij}} \\= & {} (q!)^s\left[ \frac{\beta }{q(q-1)}\right] ^{s} \sum _{i=1}^n\sum _{j=1}^n\alpha ^{\varphi _{ij}}, \end{aligned}$$

where

$$\begin{aligned} \beta = \left\{ \begin{array}{cc} \displaystyle \frac{(14q^2-4q+3)(q-1)}{24q},&{}~\text{ for } \text{ odd }~q,\\ \displaystyle \frac{(7q-2)(q-1)}{12}, &{} ~\text{ for } \text{ even }~q. \end{array}\right. \end{aligned}$$

Then, the formula (2) can be rewritten as

  1. (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)
  2. (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

$$\begin{aligned} \sum _{i=1}^n\sum _{j(\ne i)=1}^n\alpha ^{\varphi _{ij}}= & {} \sum _{i=1}^n\sum _{j(\ne i)=1}^n\text {e}^ {\eta \varphi _{ij}}\nonumber \\= & {} \sum _{i=1}^n\sum _{j(\ne i)=1}^n\sum _{t=0}^\infty \frac{\left( \eta \varphi _{ij}\right) ^t}{t!}\nonumber \\\ge & {} \displaystyle \sum _{t=0}^\infty \displaystyle \frac{p[z_{(v)}]^{t}+g[z_{(v+1)}]^{t}}{t!}\nonumber \\= & {} p\text {e}^{z_{(v)}}+g\text {e}^{z_{(v+1)}}, \end{aligned}$$
(20)

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

$$\begin{aligned} \sum _{i=1}^{n}\sum _{j(\ne i)=1}^{n}\alpha ^{\varphi _{ij}}\ge & {} p_{1}\text {e}^{z_{1(v)}}+g_{1}\text {e}^{z_{2(v+1)}}, \end{aligned}$$
(21)
$$\begin{aligned} \sum _{i=n+1}^{n+{n_1}}\sum _{j(\ne i)=n+1}^{n+{n_1}}\alpha ^{\varphi _{ij}}\ge & {} p_{2}\text {e}^{z_{2(h)}}+g_{2}\text {e}^{z_{2(h+1)}}, \end{aligned}$$
(22)
$$\begin{aligned} \sum _{i=1}^n\sum _{j=n+1}^{n+{n_1}}\alpha ^{\varphi _{ij}}\ge & {} p_{3}\text {e}^{z_{3(f)}}+g_{3}\text {e}^{z_{3(f+1)}}. \end{aligned}$$
(23)

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

$$\begin{aligned}&\left[ \overline{MD}({{\varvec{d}}}_{c})\right] ^2\nonumber \\&\quad = \left( \frac{7}{12}\right) ^{s+s_2} -2\left( \frac{7q^2+1}{12q^2}\right) ^{s+s_2} +\frac{1}{(n+n_1)^2}\left[ \sum _{i=1}^{n}\sum _{j=1}^{n} \prod _{k=1}^{s+s_2}f(x_{ik},x_{jk})\right. \nonumber \\&\qquad \left. +\sum _{i=n+1}^{n+n_1}\sum _{j=n+1}^{n+n_1} \prod _{k=1}^{s+s_2}f(x_{ik},x_{jk})+2\sum _{i=1}^{n}\sum _{j=n+1}^{n+n_1} \prod _{k=1}^{s+s_2}f(x_{ik},x_{jk})\right] \nonumber \\&\quad = \left( \frac{7}{12}\right) ^{s+s_2} -2\left( \frac{7q^2+1}{12q^2}\right) ^{s+s_2}\nonumber \\&\qquad +\frac{1}{(n+n_1)^2}\left( \frac{6q^2+1}{8q^2}\right) ^{s_2} \sum _{i=1}^{n}\sum _{j=1}^{n} \prod _{k=1}^{s}f(x_{ik},x_{jk})\nonumber \\&\qquad +\frac{1}{(n+n_1)^2}\left( \frac{14q^2-4q+3}{24q^2}\right) ^{s+s_2} \left[ \sum _{i=n+1}^{n+n_1}\sum _{j=n+1}^{n+n_1} \alpha ^{\varphi _{ij}}+2\sum _{i=1}^{n}\sum _{j=n+1}^{n+n_1} \alpha ^{\varphi _{ij}}\right] \nonumber \\&\quad = \left( \frac{7}{12}\right) ^{s+s_2} -2\left( \frac{7q^2+1}{12q^2}\right) ^{s+s_2}\nonumber \\&\qquad +\frac{1}{(n+n_1)^2}\left( \frac{6q^2+1}{8q^2}\right) ^{s_2} \sum _{i=1}^n\sum _{j=1}^n\prod _{k=1}^sf(x_{ik},x_{jk})\nonumber \\&\qquad +\frac{1}{(n+n_1)^2}\left( \frac{14q^2-4q+3}{24q^2}\right) ^{s+s_2} \left[ \sum _{i=n+1}^{n+n_1}\sum _{j(\ne i)=n+1}^{n+n_1} \alpha ^{\varphi _{ij}}+2\sum _{i=1}^{n}\sum _{j=n+1}^{n+n_1} \alpha ^{\varphi _{ij}}\right] \nonumber \\&\qquad +\frac{n_1}{(n+n_1)^2}\left( \frac{6q^2+1}{8q^2}\right) ^{s+s_2}. \end{aligned}$$
(24)

According to the formula (18), we have

$$\begin{aligned} \frac{1}{n^2}\sum _{i=1}^n\sum _{j=1}^n\prod _{k=1}^sf(x_{ik},x_{jk})= & {} [\overline{MD}({{\varvec{d}}}_0)]^2 -\left( \frac{7}{12}\right) ^s+2\left( \frac{7q^2+1}{12q^2}\right) ^s. \end{aligned}$$
(25)

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

$$\begin{aligned} \left[ \overline{MD}({{\varvec{d}}}_{c})\right] ^2= & {} \left( \frac{7}{12}\right) ^{s+s_2} -2\left( \frac{28q^2+1}{48q^2}\right) ^{s+s_2} +\frac{n_1}{(n+n_1)^2}\left( \frac{3}{4}\right) ^{s+s_2}\nonumber \\&+\frac{n^2}{(n+n_1)^2}\left( \frac{3}{4}\right) ^{s_2} \left\{ [\overline{MD}({{\varvec{d}}}_0)]^2 -\left( \frac{7}{12}\right) ^s+2\left( \frac{28q^2+1}{48q^2}\right) ^s\right\} \nonumber \\&+\frac{1}{(n+n_1)^2}\left( \frac{7q-2}{12q}\right) ^{s+s_2} \left[ \sum _{i=n+1}^{n+n_1}\sum _{j(\ne i)=n+1}^{n+n_1} \alpha ^{\varphi _{ij}}\right. \nonumber \\&\left. +2\sum _{i=1}^{n}\sum _{j=n+1}^{n+n_1} \alpha ^{\varphi _{ij}}\right] , \end{aligned}$$
(26)

which completes the proof. \(\square \)

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

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

Download citation

  • Received:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00184-020-00793-z

Keywords

Mathematics Subject Classification

Navigation