Kybernetika 52 no. 4, 575-588, 2016

D-optimal and highly D-efficient designs with non-negatively correlated observations

Krystyna Katulska and Łukasz SmagaDOI: 10.14736/kyb-2016-4-0575

Abstract:

In this paper we consider D-optimal and highly D-efficient chemical balance weighing designs. The errors are assumed to be equally non-negatively correlated and to have equal variances. Some necessary and sufficient conditions under which a design is D*-optimal design (regular D-optimal design) are proved. It is also shown that in many cases D*-optimal design does not exist. In many of those cases the designs constructed by Masaro and Wong (2008) and some new designs are shown to be highly D-efficient. Theoretical results are accompanied by numerical search, suggesting D-optimality of designs under consideration.

Keywords:

correlation, D-optimal chemical balance weighing design, Hadamard matrix, simulated annealing algorithm, tabu search, D-efficiency

Classification:

62K05, 15A18

paper.pdf

References:

  1. L. Angelis, E. Bora-Senta and C. Moyssiadis: Optimal exact experimental designs with correlated errors through a simulated annealing algorithm. Comput. Statist. Data Anal. 37 (2001), 275-296.   DOI:10.1016/s0167-9473(01)00011-1
  2. K. S. Banerjee: Weighing Designs for Chemistry, Medicine, Economics, Operations Research, Statistics. Marcel Dekker Inc., New York 1975.   CrossRef
  3. E. Bora-Senta and C. Moyssiadis: An algorithm for finding exact D- and A-optimal designs with $n$ observations and $k$ two-level factors in the presence of autocorrelated errors. J. Combinat. Math. Combinat. Comput. 30 (1999), 149-170.   CrossRef
  4. D. A. Bulutoglu and K. J. Ryan: D-optimal and near D-optimal $2^k$ fractional factorial designs of resolution $V$. J. Statist. Plann. Inference 139 (2009), 16-22.   DOI:10.1016/j.jspi.2008.05.012
  5. B. Ceranka and M. Graczyk: Optimal chemical balance weighing designs for $v+1$ objects. Kybernetika 39 (2003), 333-340.   CrossRef
  6. B. Ceranka and M. Graczyk: Robustness optimal spring balance weighing designs for estimation total weight. Kybernetika 47 (2011), 902-908.   CrossRef
  7. B. Ceranka, M. Graczyk and K. Katulska: A-optimal chemical balance weighing design with nonhomogeneity of variances of errors. Statist. Probab. Lett. 76 (2006), 653-665.   DOI:10.1016/j.spl.2005.09.012
  8. B. Ceranka, M. Graczyk and K. Katulska: On certain A-optimal chemical balance weighing design. Comput. Statist. Data Analysis 51 (2007), 5821-5827.   DOI:10.1016/j.csda.2006.10.021
  9. C. S. Cheng: Optimal biased weighing designs and two-level main-effect plans. J. Statist. Theory Practice 8 (2014), 83-99.   DOI:10.1080/15598608.2014.840520
  10. K. Domijan: tabuSearch: R based tabu search algorithm. R package version 1.1. \url{http://CRAN.R-project.org/package=tabuSearch} (2012)   CrossRef
  11. H. Ehlich: Determinantenabschätzungen für binäre Matrizen. Math. Zeitschrift 83 (1964), 123-132.   DOI:10.1007/bf01111249
  12. H. Ehlich: Determinantenabschätzungen für binäre Matrizen mit $n\equiv 3 \mathrm{mod} 4$. Math. Zeitschrift 84 (1964), 438-447.   DOI:10.1007/bf01109911
  13. Z. Galil and J. Kiefer: D-optimum weighing designs. Ann. Statist. 8 (1980), 1293-1306.   DOI:10.1214/aos/1176345202
  14. M. Graczyk: A-optimal biased spring balance weighing design. Kybernetika 47 (2011), 893-901.   CrossRef
  15. M. Graczyk: Some applications of weighing designs. Biometr. Lett. 50 (2013), 15-26.   DOI:10.2478/bile-2013-0014
  16. R. Harman, A. Bachratá and L. Filová: Construction of efficient experimental designs under multiple resource constraints. Appl. Stochast. Models in Business and Industry 32 (2015), 1, 3-17.   DOI:10.1002/asmb.2117
  17. M. Jacroux, C.S. Wong and J.C. Masaro: On the optimality of chemical balance weighing designs. J. Statist. Planning Inference 8 (1983), 231-240.   DOI:10.1016/0378-3758(83)90041-1
  18. G. M. Jenkins and J. Chanmugam: The estimation of slope when the errors are autocorrelated. J. Royal Statist. Soc., Ser. B (Statistical Methodology) 24 (1962), 199-214.   CrossRef
  19. J. S. Jung and B. J. Yum: Construction of exact D-optimal designs by tabu search. Comput. Statist. Data Analysis 21 (1996), 181-191.   DOI:10.1016/0167-9473(95)00014-3
  20. K. Katulska and Ł. Smaga: D-optimal chemical balance weighing designs with $n\equiv 0 (\text{mod} 4)$ and $3$ objects. Comm. Statist. - Theory and Methods 41 (2012), 2445-2455.   DOI:10.1080/03610926.2011.608587
  21. K. Katulska and Ł. Smaga: D-optimal chemical balance weighing designs with autoregressive errors. Metrika 76 (2013), 393-407.   DOI:10.1007/s00184-012-0394-8
  22. K. Katulska and Ł. Smaga: A note on D-optimal chemical balance weighing designs and their applications. Colloquium Biometricum 43 (2013), 37-45.   CrossRef
  23. K. Katulska and Ł. Smaga: On highly D-efficient designs with non-negatively correlated observations. REVSTAT - Statist. J. (accepted).   CrossRef
  24. C. H. Li and S. Y. Yang: On a conjecture in D-optimal designs with $n\equiv 0 (\mathrm{mod} 4)$. Linear Algebra Appl. 400 (2005), 279-290.   DOI:10.1016/j.laa.2004.11.020
  25. J. Masaro and C. S. Wong: D-optimal designs for correlated random vectors. J. Statist. Planning Inference 138 (2008), 4093-4106.   DOI:10.1016/j.jspi.2008.03.012
  26. M. G. Neubauer and R. G. Pace: D-optimal $(0,1)$-weighing designs for eight objects. Linear Algebra Appl. 432 (2010), 2634-2657.   DOI:10.1016/j.laa.2009.12.007
  27. F. Pukelsheim: Optimal Design of Experiments. John Wiley and Sons Inc., New York 1993.   CrossRef
  28. R Core Team: R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL \url{http://www.R-project.org/} (2015).   CrossRef
  29. Ł. Smaga: D-optimal Chemical Balance Weighing Designs with Various Forms of the Covariance Matrix of Random Errors. Ph.D. Thesis, Adam Mickiewicz University, 2013 (in polish).   CrossRef
  30. Ł. Smaga: Necessary and sufficient conditions in the problem of D-optimal weighing designs with autocorrelated errors. Statist. Probab. Lett. 92 (2014), 12-16.   DOI:10.1016/j.spl.2014.04.027
  31. Ł. Smaga: Uniquely E-optimal designs with $n\equiv 2 (\mathrm{mod} 4)$ correlated observations. Linear Algebra Appl. 473 (2015), 297-315.   DOI:10.1016/j.laa.2014.08.022
  32. M. Wojtas: On Hadamard's inequality for the determinants of order non-divisible by $4$. Colloquium Mathematicum 12 (1964), 73-83.   CrossRef
  33. C. H. Yang: On designs of maximal $(+1,-1)$-matrices of order $n\equiv 2 (\text{mod} 4)$. Math. Computat. 22 (1968), 174-180.   DOI:10.1090/s0025-5718-1968-0225476-4
  34. H. G. Yeh and M. N. Lo Huang: On exact D-optimal designs with $2$ two-level factors and $n$ autocorrelated observations. Metrika 61 (2005), 261-275.   DOI:10.1007/s001840400336