Optimal crossover designs in a model with self and mixed carryover effects with correlated errors (Q2256087): Difference between revisions

A crossover design is a repeated measurement design such that each experimental unit receives \(t\) different treatments during \(p\) different periods. A universally optimal design maximizes the information matrix w.r.t.\ the Loewner ordering. The authors consider crossover designs in a linear model where besides unit effects, period effects and treatment effects also carryover effects of treatments are modeled. Furthermore, the errors within each experimental unit are assumed to be correlated following a first-order autoregressive process with known parameter \(\lambda\). For this model, \textit{A. S. Hedayat} and \textit{Z.-W. Yan} [J. Stat. Plann. Inference 138, No. 7, 2201--2213 (2008; Zbl 1134.62053)] found universally optimal crossover designs for \(t \geq p = 3\). In the present, paper the authors present a close solution for \(t \geq p \geq 4\), where the proof uses the approach of \textit{H. B. Kushner} [Ann. Stat. 25, No. 6, 2328--2344 (1997; Zbl 0894.62088); correction ibid. 26, No. 5, 2081 (1998)].