Differential Weighting in Multi-Attribute Utility Measurement: When it Should Not and When it does make a Difference.

摘要

Some of the conditions under which differential weighting in multi-attribute utility measurement may be used are explored. For the case of non-negatively correlated attributes, and under conditions when no well-defined criterion variable is available, differential weighting is not important. Unit or equal weighting will do just as well. For the case of some or all of the attributes being negatively correlated and when the negative signs cannot be removed by appropriate scaling, then differential weighting can make a difference. Thus, the final choice can be quite different depending upon what weighting scheme is used. An example of choosing a 'best' automobile from a set of automobiles is given. The practical implications of these results are discussed. Decision makers may wish to retain differential weighting even for the case in which it does not make a difference, but for psychological, not numerical reasons. (Author)