Ranking alternatives - comparison of different methods based on binary comparison matrices (Q1095787)

The positive reciprocal matrix of pairwise comparisons suggested by T. L. Saaty is examined. The weights of (relative) importance of individual alternatives can be obtained by different methods. The eigenvector and logarithmic least squares methods are reviewed and two new procedures based on least squares method are introduced - one using Marquardt's method, the other using Lagrange multipliers. Comparisons of the results of all four described methods are presented, which are based on 39 sets of experimental data. The dependence of the results on the degree of inconsistency of the pairwise comparison matrix is evaluated.