D-optimal designs: an approach to reduce the number of products in preference mapping

摘要

The principle of O-optimal designs consists in choosing an optimal subset, regarding to the O-optimality criterion (D: determinant of the information matrix |X'X|), among a set of candidate points. Applied for instance to preference mapping, this approach allows, by a computing way, to select a subset of products to propose to consumers among an original set assessed quantitatively by a panel of experts. External preference mapping through the quadratic model handles then this D-optimal subset. A case study where all products are assessed by each expert panelist and each consumer is considered, allows to check interesting features of D-optimal approach in terms of minimizing the prediction error by the quadratic model of the consumer preferences on the D-optimal reduced case.