Published online by Cambridge University Press: 12 November 2009
The ever increasing availability of computer programs, personal computers, and computer facilities leaves no question that you can perform a factor analysis with little difficulty. Because of this ease, there is often a tendency to apply methods without fully understanding either the logic of the method or the more subtle question of what actually happens to the data at various stages during computation. These questions, although inextricably tied to the mathematics, can be more closely examined if one goes through the various stages in the analysis.
Few available texts on factor analysis focus their attention on these practical aspects. An exception is the book by Rummel (1970), in which factor analysis is presented as a problem in research design. We shall in this chapter outline points brought up by Rummel that are important to investigations in the natural sciences, seasoned with comments from our own experience.
The goals of any research problem must be as precisely stated as possible, for it is from these specifications that both the characteristics of the problem as well as the methodology to be employed in its resolution emanate. The analysis may involve a simple summarization or delineation of properties, or, interrelationships between properties may need to be evaluated. Another type of problem may require the study of dependencies between the variables. Still another may be concerned with associations among objects and the existence of groups in the data, thence heterogeneity.
Finally, the objectives must specify whether the results are to be used for work outside the specific study.