ABSTRACT
Network models aim at representing proximity data by means of the minimum-path-length function of connected and weighted graphs. Fundamental representation and uniqueness results underlying network models as psychological representations of stimuli, given both ordinal-scale as well as interval-scale proximity measures, are discussed. In addition, computational methods for network analyses are reviewed and compared. Methods now exist to scale metric as well as nonmetric data, symmetric and nonsymmetric proximity measures, and two-way and three-way data. They are compared with respect to the factors of (a) computational cost, (b) accuracy of recovery of an underlying network, and (c) goodness of fit to the observed proximity data.
