Designs for Generalized Linear Models With Several Variables and Model Uncertainty

摘要: Standard factorial designs sometimes may be inadequate for experiments that aim to estimate a generalized linear model, example, describing binary response in terms of several variables. A method is proposed finding exact such uses criterion allowing uncertainty the link function, predictor, or model parameters, together with design search. Designs are assessed and compared by simulation distribution efficiencies relative locally optimal over space possible models. Exact investigated two applications, their advantages central composite demonstrated.