A note on using centroid weights in additive multi-criteria decision analysis

In this short communication, the authors challenge the comparison experiments made by several authors regarding the performances of surrogate weights frequently used in additive multicriteria decision analysis. Statistical tests compare the rankings obtained, either with random 'true weights' from a given weight simplex, or with the considered surrogate techniques. It is shown that the performance ranking of the surrogate techniques entirely depends on the way the weight simplex is defined. As an example, the rank-order-centroid weights, often found in this way to be the most performant ones, are only first-ranked when a point-allocation elicitation approach is adopted. (C) 2019 Elsevier B.V. All rights reserved.