Constructions for new orthogonal arrays based on large sets of orthogonal arrays (Q6101283): Difference between revisions

An orthogonal array \(\text{OA}(N,s_{1}^{k_{1}}s_{2}^{k_{2}}\cdots s_{m}^{k_{m}},t)\) is an \(N\times k\) array with \(k=\sum_{j=1}^{m}k_{j}\) in which the first \(k_{1}\) columns have entries from \(\{0,1,2,\dots, s_{1}-1\}\), the next \(k_{2}\) columns have entries from \(\{0,1,2,\dots, s_{2}-1\}\) and so on, with the property that in any \(N\times t\) sub-array, each possible \(t\)-tuple occurs the same number of times as a row. In the special situation that \(s_{1}=s_{2}=\dots =s_{k}=s\) we say that we have a symmetric orthogonal array and denote it by \(\text{OA}(n,k,s,t)\). Orthogonal arrays were introduced by \textit{C. R. Rao} [J. R. Stat. Soc., Suppl. 9, 128--139 (1947; Zbl 0031.06201)] and are an important topic in combinatorial design theory. The parameter \(t\) is called the strength of the array. There has been quite a lot of study of orthogonal arrays of strength 2 but less of those of strength 3 or more. The paper under review gives first some results on the existence of large sets of orthogonal arrays and obtains many infinite families of them. Then it gives two effective constructions of orthogonal arrays of any strength by using large sets of orthogonal arrays and Kronecker products of matrices. Here a large set of orthogonal arrays \(\text{LOA}(N,s_{1}^{k_{1}}s_{2}^{k_{2}}\cdots s_{m}^{k_{m}},t)\) over \(S=\mathbb{Z}_{s_{1}}^{k_{1}}\times \mathbb{Z}_{s_{2}}^{k_{2}}\times \cdots \times \mathbb{Z}_{s_{m}}^{k_{m}}\) is a set \(\mathcal{L}=\{A_{1}.A_{2}.\ldots A_{M}\}\) of simple orthogonal arrays (``simple'' means no repeated row) over \(S\) with the property that every possible \(k=\sum_{j=1}^{m}k_{j}\)-tuple over \(S\) occurs in exactly one of the simple orthogonal arrays in \(\mathcal{L}\). Equivalently, the union of the \(M\) orthogonal arrays in \(\mathcal{L}\) is a trivial orthogonal array in which every possible \(k\)-tuple occurs. The authors then use Kronecker products to construct in several ways orthogonal arrays of large strength. Often the constructions in fact yield a large set of orthogonal arrays with the relevant parameters.