Abstract
In this paper, the idea of clear effects is extended to the case of blocked regular \(2^{n{-}m}\) designs with multi block variables. Some necessary and sufficient conditions on the existence of clear treatment main effects and clear treatment two-factor interactions (2fi’s) in \(2^{n{-}m}:2^l\) designs with resolution III, IV\(^- \) and IV are obtained, where the notation \(2^{n-m}:2^l\) means that there are n treatment factors, m treatment defining words and l block variables. We discuss how to find \(2^{n-m}:2^l\) designs with the maximum number of clear 2fi’s with resolution III, IV\(^-\) or IV, and present some 16-, 32-, 64-run designs containing the maximum number of clear 2fi’s in tables.
Similar content being viewed by others
References
Bisgaard S (1994) A note on the definition of resolution for blocked \(s^{n-p}\) designs. Technometrics 36:308–311
Chang JY, Notz WI (1994) Some optimal nested row-column designs. Stat Sin 4:249–263
Chen BJ, Li PF, Liu MQ, Zhang RC (2006) Some results on blocked regular 2-level fractional factorial designs with clear effects. J Stat Plan Infer 136:4436–4449
Chen H, Cheng CS (1999) Theory of optimal blocking of \(s^{n-m}\) designs. Ann Stat 27:1948–1973
Chen H, Hedayat AS (1998) \(2^{n-m}\) designs with resolution \({\rm III}\) and \({\rm IV}\) containing clear two-factor interactions. J Stat Plan Infer 75:147–158
Chen JH, Sun DX, Wu CFJ (1993) A catologue of two-level and three-level fractional factorial designs with small runs. Int Stat Rev 61:131–145
Cheng SW, Wu CFJ (2002) Choice of optimal blocking schemes in two-level and three-level designs. Technometrics 44:269–277
Łacka A, Kozłowska M, Kozłowski J (2009) Some optimal block designs with nested rows and columns for research on alternative methods of limiting slun damage. Stat Papers 50:837–846
Sitter RR, Chen J, Feder M (1997) Fractional resolution and minimum aberration in blocked \(s^{n-k}\) designs. Technometrics 39:382–390
Wei JL, Zhang RC (2014) Blocked two-level regular designs with general minimum lower-order confounding. J Stat Theory Pract 8:46–65
Wu CFJ, Hamada MS (2009) Expeiments: planning, analysis, and optimization. Wiley, New York
Wu H, Mee R, Tang B (2012) Fractional factorial designs with admissible sets of clear two-factor interactions. Technometrics 54:191–197
Xu HQ (2009) Algorithmic construction of efficient fractional factorial designs with large run sizes. Technometrics 51:262–277
Zhang RC, Li P, Wei JL (2011) Optimal two-level regular designs with multi block variables. J Stat Theory Pract 5:161–178
Zhang RC, Li P, Zhao SL, Ai MY (2008) A general minimum lower-order confounding criterion for two-level regular designs. Stat Sin 18:1689–1705
Zhang RC, Mukerjee R (2009) General minimum lower-order confounding in block designs using complementary sets. Stat Sin 19:1787–1802
Zhang RC, Park DK (2000) Optimal blocking of two-level fractional factorial designs. J Stat Plan Infer 91:107–121
Zhao SL, Li PF (2016) Construction of minimum aberration blocked two-level regular factorial designs. Commun Stat Theory Methods 45(17):5028–5036
Zhao SL, Li PF, Karunamuni R (2013) Blocked two-level regular factorial designs with weak minimum aberration. Biometrika 100:249–253
Zhao SL, Li PF, Liu MQ (2013) On blocked resolution IV designs containing clear two-factor interactions. J Complex 29:389–395
Zhao SL, Li PF, Zhang RC, Karunamuni R (2013) Construction of blocked two-level regular factorial designs with general minimum lower-order confounding. J Stat Plan Infer 143:1082–1090
Zhao SL, Sun Q (2017) On constructing general minimum lower order confounding two-level block designs. Comnum Stat Theory Methods 46(3):1261–1274
Zhao YN, Zhao SL, Liu MQ (2016) A theory on constructing blocked two-level designs with general minimum lower-order confounding. Front Math China 11:207–235
Zhao YN, Zhao SL, Liu MQ (2017) On constructing optimal two-level designs with multi block variables. J Syst Sci Complex, accepted
Acknowledgements
The authors would like to thank the editor and the referees for the constructive suggestions and comments that lead to a significant improvement over the article. This work was partially supported by the National Natural Science Foundation of China Grant 11371223; the Program for Scientific Research Innovation Team in Applied Probability and Statistics of Qufu Normal University (No. 0230518); and the Scientific and Technological Program of Qufu Normal University (No. xkj201519).
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
In the following tables, \(\mathbf {1,\ldots ,k}\) are the independent columns defined in (1) and the other columns are component-wise products of the k independent columns. Each design contains the independent columns \(\mathbf {1,\ldots ,k}\) and m additional columns of \(H_k\), where \(k=n-m\), and only the additional m columns of the \(2^{n-m}\) design \(D_0\) are shown in the tables under the column heading Additional columns. BC represents the number of clear 2fi’s in the \(2^{n-m}:2^l\) designs, and for all the different n, k and l, \(*=2^k-1-n-l\).
Tables 1 and 2 present the 16- and 32-run \(2_\mathrm{III}^{n-m}:2^l\) designs, containing the maximum number of clear 2fi’s, respectively. In Table 1, take the \(2^{6-2}:2^l\) design \(D=(D_0,B)\) for example, where \(D_0=(\mathbf {1},\mathbf {2},\mathbf {3},\mathbf {4},\mathbf {12},\mathbf {34})\). For any \(1\le l\le 8\), under the column heading Block variables, \(b_1,\ldots ,b_l\in H_4\backslash D_0\) means that we can select any l columns from \(H_4\backslash D_0\) as block variables B of the \(2^{6-2}:2^l\) design. In Table 2, take the \(2^{7-2}:2^l\) design \(D=(D_0,B)\) for example, where \(D_0=(\mathbf {1},\mathbf {2},\mathbf {3},\mathbf {4},\mathbf {5},\mathbf {12},\mathbf {1345})\). For \(1\le l\le 6\), any l columns of \(\{\mathbf {234},\mathbf {235},\mathbf {245},\mathbf {1234},\mathbf {1235},\mathbf {1245}\}\) under the column heading Block variables make the block variables B; for \(7\le l\le 23\), under the column heading Block variables, other\(b_i\in H_5\backslash D_0, 7\le i\le l\) means that we take all the 6 columns of \(\{\mathbf {234},\mathbf {235},\mathbf {245},\mathbf {1234},\mathbf {1235},\mathbf {1245}\}\) and any other \(l-6\) columns from \(H_5\backslash D_0\) as block variables B. In the two tables, the \(2^{8-4}:2^l\) design and the \(2^{10-5}:2^l\) design are similar to the \(2^{6-2}:2^l\) design. The other blocked designs are similar to the \(2^{7-2}:2^l\) design.
In Tables 3 and 4, under the column heading \(n-m\), \(n-m.j\) denotes the jth \(2^{n-m}\) design in the catalogue. For each given \(\{n, m, l\}\), the first \(2^{n-m}:2^l\) design \(D=(D_0,B)\) is the blocked design containing the maximum number of clear 2fi’s with resolution at least IV\(^-\), where \(D_0\) is determined by \(n-m.1\) under the column heading \(n-m\) and B is determined by the corresponding l block variables under the column heading Block variables. These blocked designs are constructed by the method in Section 3. By the algorithm of Section 4, we obtain the \(2_\mathrm{IV}^{n-m}:2^l\) design containing the maximum number of clear 2fi’s. For each \(2_\mathrm{IV}^{n-m}:2^l\) design, all the columns of the \(\varphi \)-class are listed under the column heading Block variables. For any given l, we can choose any l columns from \(\varphi \)-class to form the block variables. Take the \(2^{7-2}:2^l\) design \(D=(D_0,B)\) for example, where \(D_0=(\mathbf {1},\mathbf {2},\mathbf {3},\mathbf {4},\mathbf {5},\mathbf {123},\mathbf {1245})\) and the \(\varphi \)-class of \(D_0\) is \(\{\mathbf {134},\mathbf {234},\mathbf {135},\mathbf {235},\mathbf {1345},\mathbf {2345}\}\). When \(1\le l\le 6\), we can choose any l columns from the \(\varphi \)-class as block variables; thus we obtain the \(2^{7-2}:2^l\) design \(D=(D_0,B)\), which contains 15 clear 2fi’s, the maximum number of clear 2fi’s among all the \(2_\mathrm{IV}^{7-2}:2^l\) designs. When \(7\le l\le 9\), \(b_7,\ldots ,b_l\in UC\)-class means that we choose all the 6 columns of the \(\varphi \)-class and any \(l-6\) columns from the UC-classs as block variables; such a \(2^{7-2}:2^l\) design has resolution IV\(^-\) and 15 clear 2fi’s. When \(10\le l\le 23\), other \(b_i\in C\)-class, \(10\le i\le l\) means that we choose all the 9 columns of the \(\varphi \)-class and the UC-class and any \(l-9\) columns from the C-class as block variables; such a \(2^{7-2}:2^l\) design has resolution IV\(^-\) and \(2^k-1-n-l=24-l\) clear 2fi’s.
Rights and permissions
About this article
Cite this article
Zhao, Q., Zhao, S. Some results on two-level regular designs with multi block variables containing clear effects. Stat Papers 60, 1569–1582 (2019). https://doi.org/10.1007/s00362-017-0886-8
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00362-017-0886-8