Mapping Associations - Mapping Scientific Frontiers: The Quest for Knowledge Visualization - page 120

Fig. 3.31 SCI multidimensional scaling display with cluster boundaries (Reproduced from Morris
and McCain ( 1998 ). Note the distance between “Comput Biol Med” and “Int J Clin Monit Comput”
to the left of this MDS configuration)
measures between the p objects. The model explains differences between subjects'
cognition by a variant of the distance model. The p objects are represented as points
in a space known as a master space, a shared space, or a group space. The subjects
perceive this space differently because individuals afford a different salience or
weight to each dimension of the space. The INDSCAL model assumes that subjects
are systematically distorting the group space and it seeks to reconstruct both the
individual private, distorted spaces and the aggregate “group” space. Similarity
measures can be derived from aggregated groups as well as from individuals'
ratings. For example, in judging the differences between two houses an architect
might primarily concentrate on style and structure, whereas a buyer might be more
concerned with the difference in price.
Carroll and Chang illustrated INDSCAL with an example of analyzing how
people perceive the distances between six different areas of a city. They asked three
subjects to estimate the distance between each of the pairs of areas. Each subject
estimated a total of 15 such pairs, (6 5)/2 D 15.
The INDSCAL model interprets individual differences in terms of subjects
applying individual sets of weights to the dimension of a common “group” or