RUS  ENG
Full version
JOURNALS // Upravlenie Bol'shimi Sistemami // Archive

UBS, 2023 Issue 103, Pages 135–170 (Mi ubs1154)

Control in Social and Economic Systems

Methods for selecting the median ranking and evaluating the consistency of expert assessments by the proximity criterion

V. P. Korneenko

V.A. Trapeznikov Institute of Control Sciences of RAS, Moscow

Abstract: In the development of the theory of expert assessments, the exceptional role of the position and the median of ranking, known as the Kemeny median, has been revealed. However, there is no optimal solution method for finding the median of the rankings represented by the matrices of binary relations according to the distance matrix criterion. The validity of the optimal solution to the problem of choosing the median in the space of the rank scale of measurement is due to the fact that there is a one-to-one correspondence between the rankings represented by binary relation matrices on a set of pairs of objects and the rankings in the rank scale. It is also an important task to check the consistency of the opinions of the expert group. The existing statistical methods and methods of rank correlation do not measure the consistency of expert opinions, if by which we mean the measure of proximity between expert assessments of objects.. The article shows by concrete examples that the Kendal concordance coefficient, which is still found in the works of some authors, does not allow for a realistic assessment of the consistency of expert rankings, which can lead to erroneous management decisions. A method is proposed for evaluating the opinions of both a pair of experts and a group of experts, in the form of an average agreement of experts with respect to the median of rankings presented in the ranking scale.

Keywords: rank scale, related (rational) ranks, median ranking of objects, matrix criterion, consistency of expert opinions.

UDC: 519.8
BBK: 22.18

Received: March 27, 2023
Published: May 31, 2023

DOI: 10.25728/ubs.2023.103.6



© Steklov Math. Inst. of RAS, 2026