Abstract
Computer experiments usually involve many factors, but only a few of them are active. In such a case, it is desirable to construct designs with good projection properties. Maximum projection designs and uniform projection designs have been developed for common experimental situations, however, there has been little study on constructing projection designs for high-accuracy computer experiments (HEs) and low-accuracy computer experiments (LEs) so far. In this paper, we propose a weighted uniform projection criterion, and construct uniform projection nested Latin hypercube designs to suit such computer experiment situations. We show that the obtained designs have good projection properties in all sub-dimensions, and we also discuss how to choose a proper value for the weight. Simulated examples are available to illustrate the effectiveness of the proposed designs.
Similar content being viewed by others
Notes
In the DACE, we set the parameters \(theta0=[5,5,5,5,5]\), \(lob=[0.01,0.01,0.01,0.01,0.01]\) \(upb=[20,20,20,20,20]\).
References
Ba S, Myers WR, Wang DP (2018) A sequential maximum projection design framework for computer experiments with inert factors. Stat Sin 28:879–897
Boukouvalas A, Gosling JP, Maruri-Aguilar H (2014) An efficient screening method for computer experiments. Technometrics 56:422–431
Chen H, Liu MQ (2015) Nested Latin hypercube designs with sliced structures. Commun Simul Theory Methods 44:4721–4733
Chen D, Xiong S (2017) Flexible nested Latin hypercube designs for computer experiments. J Qual Technol 49:337–353
Chen H, Huang HZ, Dennis LKJ, Liu MQ (2016) Uniform sliced Latin hypercube designs. Appl Stoch Models Bus Ind 32:574–584
Fang KT, Li R, Sudjianto A (2006) Design and modeling for computer experiments. CRC Press, New York
Fang KT, Liu MQ, Qin H, Zhou YD (2018) Theory and application of uniform experimental designs. Springer and Science Press, Singapore and Beijing
Haaland B, Qian PZG (2010) An approach to constructing nested space-filling designs for multi-fidelity computer experiments. Stat Sin 20:1063–1075
Han G, Santner TJ, Notz WI, Bartel DL (2009) Prediction for computer experiments having quantitative and qualitative input variables. Technometrics 51:278–288
Hickernell FJ (1998) A generalized discrepancy and quadrature error bound. Math Comput 67:299–322
Jin R, Chen W, Sudjianto A (2005) An efficient algorithm for constructing optimal design of computer experiments. J Stat Plan Inference 134:268–287
Joseph VR, Gul E, Ba S (2015) Maximum projection designs for computer experiments. Biometrika 102:371–380
Kennedy MC, O’Hagan A (2000) Predicting the output from a complex computer code when fast approximations are available. Biometrika 87:1–13
Lophaven SN, Nielsen HB, Sondergaard J (2002) A Matlab kriging toolbox DACE. Version 2.5. http://www2.imm.dtu.dk/pubdb/p.php?1460
McKay MD, Beckman RJ, Conover WJ (1979) A comparison of three methods for selecting values of input variables in the analysis of output from a computer code. Technometrics 21:239–245
Moon H, Dean A, Santner T (2011) Algorithms for generating maximin Latin hypercube and orthogonal designs. J Stat Theory Pract 5:81–98
Morris MD, Mitchell T (1995) Exploratory designs for computational experiments. J Stat Plan Inference 43:381–402
Qian PZG (2009) Nested Latin hypercube designs. Biometrika 96:957–970
Qian PZG, Wu CFJ (2008) Bayesian hierarchical modeling for integrating low-accuracy and high-accuracy experiments. Technometrics 50:192–204
Qian PZG, Seepersad C, Joseph R, Allen J, Wu CFJ (2006) Building surrogate models based on detailed and approximate simulations. ASME Trans J Mech Des 128:668–677
Qian PZG, Wu H, Wu CFJ (2008) Gaussian process models for computer experiments with qualitative and quantitative factors. Technometrics 50:383–396
Qian PZG, Ai MY, Wu CFJ (2009a) Construction of nested space-filling designs. Ann Stat 37:3616–3643
Qian PZG, Tang B, Wu CFJ (2009b) Nested space-filling designs for computer experiments with two levels of accuracy. Stat Sin 19:287–300
Reese CS, Wilson AG, Hamada M, Martz HF, Ryan KJ (2004) Integrated analysis of computer and physical experiments. Technometrics 46:153–164
Rennen G, Husslage BGM, Van Dam ER, Den Hertog D (2010) Nested maximin Latin hypercube designs. Struct Multidiscipl Optim 41:371–395
Saab YG, Rao YB (1991) Combinational optimization by stochastic evolution. IEEE Trans Comput Aided Des 10:525
Santner TJ, Williams BJ, Notz WI (2003) The design and analysis of computer experiments. Springer, New York
Schmidt RR, Cruz EE, Iyengar MK (2005) Challenges of data center thermal management. IBM J Res Dev 49:709–723
Sun FS, Wang YP, Xu H (2019) Uniform projection designs. Ann Stat 47:641–661
Sun FS, Yin YH, Liu MQ (2013) Construction of nested space-filling designs with two levels of accuracy using difference matrices. J Stat Plan Inference 143:160–166
Van Dam ER, Husslage BGM, Den Herttog D (2004) One-dimensional nested maximin designs. (CentER Discussion Paper; Vol. 2004-66). Tilburg: Operations research
Xu J, Duan XJ, Wang ZM, Yan L (2018) A general construction for nested Latin hypercube designs. Stat Prob Lett 134:134–140
Yang JY, Liu MQ, Lin DKJ (2014) Construction of nested orthogonal Latin hypercube designs. Stat Sin 24:211–219
Acknowledgements
The authors thank the Editor-in-Chief and two reviewers for their valuable comments and suggestions. This work was supported by the National Natural Science Foundation of China (Grant Nos. 11601367, 11601366, 11771219), and the ‘131’ Talent Program of Tianjin. The first two authors contributed equally to this work.
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.
Appendices
Appendix
Proof of Theorem 1
An LHD \(D=L(n,q)\) is a balanced design, thus it satisfies the condition of Theorem 3 in Sun et al. (2019). Note that the LB of \(\phi (D)\) in Sun et al. (2019) is obtained based on squared centered \(L_2\)-discrepancy, denoted by CD(D), then \(CD_2(D)=\sqrt{CD(D)}\). For the criterion (1), we have
This completes the proof of Theorem 1.
To prove Theorem 2, we need the conclusion of the Theorem 1 in Sun et al. (2019). For more details, please refer to the paper.
Proof of Theorem 2
From the proof of Theorem 1 in Sun et al. (2019), we have
for any \(2 \le r \le q\), even if there is no balanced condition for the design D. Thus for both \(D_h\) and \(D_l\) in an UPNLHD, Eq. (12) holds as well. According to the definition of criterion in (4), we have
where \(D_{h,u}\) is the projected deign of \(D_h\) onto r-dimensional space indexed by u for any \(2\le r \le q\), and \(D_{l,u}\) is similarly defined. This completes the proof of Theorem 2.
Rights and permissions
About this article
Cite this article
Chen, H., Zhang, Y. & Yang, X. Uniform projection nested Latin hypercube designs. Stat Papers 62, 2031–2045 (2021). https://doi.org/10.1007/s00362-020-01172-6
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00362-020-01172-6