This short course will examine multidimensional scaling (MDS), with particular attention to its utility as a research tool for analyzing public opinion data. The basic objective of MDS is to produce a geometric model of stimuli, based upon information about the Aproximities among those stimuli. In this model, each stimulus is shown as a point within a space. The distances between the points correspond to the proximities between the stimuli. That is, stimuli that are more Aproximal to each other will be shown as points that are located relatively close to each other within the space. Less proximal stimuli are shown as points that are located farther apart from each other.

Multidimensional scaling methods have many potential applications in public opinion research. They are commonly used to estimate the cognitive structures underlying survey responses, using respondents' perceptions of the similarities among a set of objects (e.g., political candidates, issue positions, and so on). MDS can be adapted to represent respondent preferences among a set of stimuli. It can also be generalized to show individual differences across subsets of respondents.

MDS can be employed with many kinds of input data. And nonmetric MDS only requires ordinal-level information; hence, it is ideally suited for the kinds of data obtained from public opinion surveys. The fact that MDS produce metric (i.e., interval-level) output means that it is potentially very useful as a measurement tool.

The content of the short course will rely on intuition and graphical displays rather than rigorous mathematical or statistical treatments. It will use an ongoing, easy-to-understand, example (drawn from the field of public opinion) to explain and illustrate the methodology. We will cover the following

The basic idea of MDS; the general estimation procedure; interpretation of output; different varieties of MDS; and software for performing MDS (although the examples used in the short course will rely primarily upon SPSS). Participants in the course should be able to employ MDS techniques immediately in their own research.

COST: $25 students, $50 nonstudents; discount when paid with conference registration