Limits and extensions of equal weights in additive multiattribute models

Abstract The equal weights is a dominant idea in psychology is implicit in Wainer's salient title. ‘Estimating coefficients in linear models: It don't make no nevermind’. The claim for equal weights should be more modest. In this paper it will be demonstrated, that in selecting a single best multiattributed alternative, the weights do matter. The procedure reported determines ‘nearly equal’ weights which solve (subject to obvious feasibility requirements) the following two problems: determine the most ‘nearly equal’ weights which 1. equate the overall multiattribute value of the best alternative based on equal weights, and any other non-dominated alternative; 2. reverse the overall multiattribute value of the best ‘equal weights’ alternative and any other non-dominated alternative by a pre-specified difference in value, thereby inducing a loss in multiattribute value. As a sensitivity analysis tool, the procedure leads to a synthesis of man and model which can improve overall judgment. An example using 15 alternatives of nine attributes each illustrates the procedure(s).