| Abstract: | Many situations exist in which multiple dimensions must be combined to create a composite which reflects overall utility or value to a decision maker. A linear combination is often used in which the decision maker's importance weights are multiplied by values of each dimension and then added. When the decision maker's weights are applied, however, the dimension values must have equal variance and essentially zero correlations, or else the ordering on the composite will be incorrect. In this paper, we provide a method which compensates for dimension correlation and inequality of dimension variance. The method is illustrated with ratings of modifications proposed for nuclear power plants. |
