Optimal experimental design procedures are useful for quantitative responses under nonstandard conditions. These methods, however, require prior knowledge of the form of the response function, which is often not available. This paper develops a model-robust technique, i.e., one which is efficient to a set of possible models, connected to multiresponse D-optimal design. The model-robust design is constructed using a generalization of the modified Federov exchange algorithm. The paper also compares designs robust for large sets of models with those robust for smaller sets, including pairs.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados