Increasing Convergence of the Quality of Estimated Interval Weight Vector in Interval AHP

Abstract

As one of desirable properties of an interval priority weight estimation method, we consider the increasing convergence of the quality of estimated interval priority weights with respect to the number of observations of the pairwise comparison matrix. This convergence implies that the quality of the estimated interval priority weights is improved and converges to a certain value as the preference information obtained from the decision maker increases. In this paper, by a numerical experiment, we show that previously proposed interval priority weight estimation methods do not possess this convergence property although the initial estimation quality is sufficiently good. Then we propose several interval priority weight estimation methods having this convergence property as well as a sufficiently good initial estimation quality.