METRIC VS. NONMETRIC PROCEDURES FOR MULTIATTRIBUTE MODELING: SOME SIMULATION RESULTS *

Abstract
Nonmetric procedures such a MONANOVA have been developed to estimate attribute utilities with ranked observations. It is argued that the goodness of the estimation procedure is another criterion which favors a metric procedure like Ordinary Least Squares (OLS) Regression. MONANOVA and OLS Regression are compared in their ability to recover attribute utilities with both ranked and scaled observations.