Minimax Designs for Approximately Linear Multiresponse Models

This paper considers the minimax design problem in approximately linear multiresponse regression models. The difference between the fitted linear models and the true responses is a vector of non-linear contaminations restricted by a bound on its L_2 norm in the design region. We derive a theory to guide in the construction of designs which minimize the maximum loss. The minimax design is found for a bivariate regression model with two responses.