Factor Analysis

Exploratory factor analysis is a method for explaining the covariance relationships amongst a number of observed variables in terms of a much smaller number of unobserved variables, termed factors.

EViews provides a wide range of tools for performing factor analysis, from computing the covariance matrix from raw data all the way through the construction of factor score estimates.

Factor analysis in EViews is carried out using the factor object. The remainder of this chapter describes the use of the EViews factor object to perform exploratory factor analysis. Using the EViews factor object you may:

• Compute covariances, correlations, or other measures of association.

• Specify the number of factors.

• Obtain initial uniqueness estimates.

• Extract (estimate) factor loadings and uniquenesses.

• Examine diagnostics.

• Perform factor rotation.

• Estimate factor scores.

EViews provides a wide range of choices in each of these areas. You may, for example, select from a menu of automatic methods for choosing the number of factors to be retained, or you may specify an arbitrary number of factors. You may estimate your model using principal factors, iterated principal factors, maximum likelihood, unweighted least squares, generalized least squares, and noniterative partitioned covariance estimation (PACE). Once you obtain initial estimates, rotations may be performed using any of more than 30 orthogonal and oblique methods, and factor scores may be estimated in more than a dozen ways.

We begin with a discussion of the process of creating and specifying a factor object and using the object to estimate the model, perform factor rotation, and estimate factor scores. This section assumes some familiarity with the common factor model and the various issues associated with specification, rotation, and scoring. Those requiring additional detail may wish to consult
“Background”.

Next, we provide an overview of the views, procedures, and data members provided by the factor object, followed by an extended example highlighting selected features.

The remainder of the chapter provides relevant background information on the common factor model. Our discussion is necessarily limited; the literature on factor analysis is extensive, to say the least, and we cannot possibly attempt a comprehensive overview. For those requiring a detailed treatment, Harman’s (1976) book length treatment is a standard reference. Other useful surveys include Gorsuch (1983) and Tucker and MacCallum (1977).