Cluster Analysis, Variables
Ralph B. D’Agostino, Heidy K. Russell, Taiyeong Lee · Wiley StatsRef: Statistics Reference Online · 2017
Abstract Cluster analysis involves grouping objects, subjects or variables, with similar characteristics into groups. Similarity or dissimilarity of objects is measured by a particular index of association. The focus here is on clustering of variables instead of subjects. Types of methods that cluster variables based on correlation structure of variables or factorial structure of variables are considered, where factorial structure refers to a structure obtained from principal component analysis or factor analysis. The approaches described for cluster analysis use various indices of association such as factor loadings, correlation matrix, and cosine matrix. Examples of these are shown and the advantages and disadvantages of each approach are discussed.