Multidimensional Scaling

Given a set of objects, the goal of multidimensional scaling is to find a representation of the objects in a low-dimensional space. This solution is found by using the proximities between the objects. The procedure minimizes the squared deviations between the original, possibly transformed, object proximities and their Euclidean distances in the low-dimensional space.

The purpose of the low-dimensional space is to uncover relationships between the objects. By restricting the solution to be a linear combination of independent variables, you may be able to interpret the dimensions of the solution in terms of these variables. In the following example, you will see how 15 different kinship terms can be represented in three dimensions and how that space can be interpreted with respect to the gender, generation, and degree of separation of each of the terms.

Next