Blackwell Reference Online

Extract

Factor analysis is a statistical method for empirically identifying the structure underlying measured or factored entities (e.g., variables). The three purposes for which factor analysis can be used are (1) empirically creating a theory of structure (e.g., Cattell's Structure of Intellect model), (2) evaluating whether factored entities (e.g., variables) cluster in a theoretically expected way (e.g., construct validity evaluation), and (3) estimating latent variables scores (i.e., factor scores) that are then used in subsequent statistical analyses (e.g., MANOVA, descriptive discriminant analysis) in place of the measured factored entities (e.g., variables). In common analytic practice, the factored entities are usually (1) variables, although (2) people and (3) occasions of measurement also can be factored ( Thompson 2000 ). Factor analysis statistical software does not know if it is factoring variables, people, or time, and from a statistical point of view, the mathematics of factor analysis can sensibly be invoked for any of these possibilities. The data matrix for the analysis is created such that the entities to be factored (e.g., variables) constitute the columns of the matrix. The rows constitute the dimension over which patterns of association (e.g., correlation, covariance) among the factored entities are estimated. The factors are then estimated based on these association ... log in or subscribe to read full text