Optimality aspects of 3-concurrence most balanced designs (Q1115070): Difference between revisions

\textit{K. Takeuchi} [Rep. Stat. Appl. Res. Union Jap. Sci. Eng. 8, 140-145 (1961; Zbl 0104.125)] established E-optimality of Group Divisible Designs (GDDs) with \(\lambda_ 2=\lambda_ 1+1\). Much later, \textit{C.-S. Cheng} [J. R. Stat. Soc. Ser. B 42, 199-204 (1980; Zbl 0433.62047)] and \textit{M. Jacroux} [ibid. 44, 205-209 (1980; Zbl 0443.62063)] demonstrated E-optimality property of the GDDs with \(n=2\), \(\lambda_ 1=\lambda_ 2+1\) or with \(m=2\), \(\lambda_ 2=\lambda_ 1+2\). The purpose of this paper is to provide a unified approach for identifying certain classes of designs as E-optimal. In the process, we come up with a complete characterization of all E-optimal designs attaining a specific bound for the smallest non-zero eigenvalue of the underlying C-matrices. This establishes E-optimality of a class of 3-concurrence most balanced designs with suitable intra- and inter-group balancing. We also discuss the MV-optimality aspect of such designs.