SPSS Statistics for Data Analysis and Visualization

Chapter 10 Display Complex Relationships with Multidimensional Scaling

Multidimensional scaling (MDS) is an exploratory statistical technique that positions objects (products, services, and so on) in a graph based on how similar or dissimilar the objects are to each other so as to determine which features are most important in people’s perceptions of object relatedness. Multidimensional scaling mathematically transforms the perceived relatedness among objects into a visual display of distance that helps determine which characteristics are most important in detecting the structure or dimensions of similarity judgments. Dimensions are then interpreted by using both subjective and statistical techniques.

As an example, in the area of market research, multidimensional scaling can be used to identify brand positions and the dimensions along which customers view brands. In its most basic form, respondents indicate how similar or dissimilar different pairs of objects (often brands) are from one another. Using ice cream as an example, we can ask respondents, how similar are vanilla and chocolate? Chocolate and strawberry? Vanilla and strawberry? Lemon and raspberry? Multidimensional scaling then attempts to fit these proximity measures by placing the objects in a low-dimensional space so that the distances between the objects in the space reproduce the observed proximities.

Figure 10.1 shows a two dimensional display of the relationships among ice cream flavors. Notice that the first ...