When to combine group judgments and when not to in the analytic hierarchy process: A new method

In paired comparisons, the individuals in a number of groups giving their judgments are examined for agreement among themselves. Statistical methods are applied to see whether these groups are significantly different in giving their judgments or not. When individual judgments are homogeneous, a single matrix is obtained for all of them to obtain a common rank order and weights. Otherwise, separate matrices are estimated for significantly different group judgments. Two numerical examples are given to explain the methodologies developed in the paper.