Ranking decision variants by subjective paired comparisons in cases with incomplete data

Abstract

The pairwise comparison method is widely used to rank a finite, usually small number of decision variants especially in a case when neither a direct evaluation nor the utility theory gives satisfactory results. In this method, an expert or a group of experts is asked to provide his/their opinions concerning each pair of factors expressing a relative importance of one variant in a pair over the second one. It happens however that an expert or few experts cannot provide his/their opinions concerning a pair or pairs of factors. In such a case the resulting judgement matrices are incomplete and a problem of estimating missing data arises. The paper addresses some issues concerning lacking data. Some numerical examples are included. © Springer-Verlag 2003.