Decision making during consistent expert estimations of pairwise comparisonsDecision making during consistent expert estimations of pairwise comparisons

Abstract

We consider two groups of methods for pairwise comparisons to compute the relative weights of alternative solutions based on expert estimations: the "line" and "triangle" methods. In the "line" methods, the weights of n alternatives are calculated based on n-1 expert estimations of pairwise comparisons made in the scale and assumed full consistency of expert knowledge. The "triangle" methods to calculate weights require an excessive number of n(n-1)/2 expert estimations, which are used to evaluate the consistency of expert knowledge. The results obtained by the two groups of methods were compared. Using a simulation of high expert competence in assessing the alternative solutions by paired comparisons methods under the Saaty scale, the weight errors estimations were calculated using the "triangle" and "line" methods. It is shown that the condition of a complete consistency of expert estimations of paired comparisons, made under the scale, can introduce an additional error during the construction of the paired comparisons matrix and, consequently, into the resulting weights.