Cluster Analysis

Cluster analysis intends to provide groupings of set of items, objects, or behaviors that are similar to each other. The outcome of a cluster analysis provides the set of associations that exist among and between various groupings that are provided by the analysis. The actual technique depends on the application of multivariate statistics and the generation of association from available inputs. The choice of input variables becomes important in both providing the basis for similarity within a grouping and differentiating the differences between groupings.

For communication, suppose a research goal is to cluster public speakers on the basis of various behaviors exhibited during a presentation. So, a variety of behaviors are rated (e.g., speaking rate in words per minute, number of hand gestures, number of steps taken, number of shrugs of shoulders, number of disfluencies, length of pauses used during the presentation, number of transitions provided, and number of internal summaries used). Notice that the clustering does not provide which speakers would be considered strong or weak; the cluster is simply on the basis of similarity of behaviors. Thus, the clustering provides relations among various elements and the degree to which those elements are related. The remainder of this entry considers an application of the example of public speaking to illustrate the various approaches and uses of cluster analysis.