Optimal designs for mixed continuous and binary responses with quantitative and qualitative factors (Q2222236)

The authors consider a multivariate regression model of mixed variable types, continuous and binary, and on quantitative and qualitative factors. As usual in the theory of optimal designs, they construct first the information matrix. They ask for local \(\Phi_p\)-optimal designs. For this, they adapt the complete class approach by \textit{M. Yang} and \textit{J. Stufken} [Ann. Stat. 40, No. 3, 1665--1681 (2012; Zbl 1257.62083)]. As a further key ingredient, they use vectors of functions that form a Chebyshev system. The complete class results lead to a small class of designs within which an optimal design can be found by well-developed optimization techniques, e.g. by sequential quadratic programming. The results are applied to a model for a breast cancer study which uses logistic regression for the binary response.