Abstract
In this paper, I continue to study the alternative hierarchical analysis method that was initialed in [Saaty, 1976]. Instead of using pairwise ratio matrices, this method uses pairwise subtraction matrices. By doing so, computational complexity is reduced significantly, and as shown by a statistical experiment described in this paper, the two methods lead basically to the same results. Furthermore, this method provides an optimal solution in the sense of minimizing an appropriate error function. Due to the significant reduction in computational complexity, fuzzy hierarchical analysis based on pairwise subtraction matrices becomes more manageable. Finally, I point out a new application area of the hierarchical analysis: estimating crisp or fuzzy prior probability distributions for Bayesian inference with imprecise probabilities.