On optimality of designs with three distinct eigenvalues (Q1953495): Difference between revisions

Summary: Let \({\mathcal{D}}_{v,b,k}\) denote the family of all connected block designs with \(v\) treatments and \(b\) blocks of size \(k\). Let \(d\in{\mathcal{D}}_{v,b,k}\). The replication of a treatment is the number of times it appears in the blocks of \(d\). The matrix \(C(d) = R(d)-\frac{1}{k}N(d)N(d)^\top\) is called the information matrix of \(d\) where \(N(d)\) is the incidence matrix of \(d\) and \(R(d)\) is a diagonal matrix of the replications. Since \(d\) is connected, \(C(d)\) has \(v-1\) nonzero eigenvalues \(\mu_1(d),\ldots,\mu_{v-1}(d)\).Let \({\mathcal{D}}\) be the class of all binary designs of \({\mathcal{D}}_{v,b,k}\). We prove that if there is a design \(d^*\in{\mathcal{D}}\) such that {\parindent=7mm \begin{itemize}\item[(i)]\(C(d^*)\) has three distinct eigenvalues, \item[(ii)]\(d^*\) minimizes trace of \(C(d)^2\) over \(d\in{\mathcal{D}}\), \item[(iii)]\(d^*\) maximizes the smallest nonzero eigenvalue and the product of the nonzero eigenvalues of \(C(d)\) over \(d\in{\mathcal{D}}\), then for all \(p>0\), \(d^*\) minimizes \(\left(\sum_{i=1}^{v-1}\mu_i(d)^{-p}\right)^{1/p}\) over \(d\in{\mathcal{D}}\). \end{itemize}} In the context of optimal design theory, this means that if there is a design \(d^*\in{\mathcal{D}}\) such that its information matrix has three distinct eigenvalues satisfying the condition (ii) above and that \(d^*\) is E- and D-optimal in \({\mathcal{D}}\), then \(d^*\) is \(\Phi_p\)-optimal in \({\mathcal{D}}\) for all \(p>0\). As an application, we demonstrate the \(\Phi_p\)-optimality of certain group divisible designs. Our proof is based on the method of KKT conditions in nonlinear programming.