ABSTRACT

Multidimensional scaling is one of several multivariate techniques that aim to reveal the structure of a data set by plotting points in one or two dimensions. The basic idea can be motivated by a geographical example. Suppose we are given the distances between pairs of cities and are asked to reconstruct the two-dimensional map from which those distances were derived. We could attempt to do this by a process of trial and error by moving points about on a sheet of paper until we got the distances right. A procedure that does this automatically is called multidimensional scaling (MDS). The “multi” part of the name refers to the fact that we are not restricted to constructing maps in one or two dimensions.