Factor Analysis, Information-Transforming Instruments, and Objectivity: A Reply and Discussion

Abstract

Davis Baird [1987] has called attention to a certain way of looking at statistical procedures used in the behavioral and social sciences, such as exploratory common factor analysis: they are information-transforming instruments. He believes that looking at exploratory factor analysis and other statistical procedures in this way renders incoherent any critique of them (such as mine, Mulaik [1985], [1988]) as, say, knowledge discovering procedures (a rather strong view one might have about these procedures) or as optimal hypothesis generating procedures (a somewhat more cautious view). Exploratory factor analysis and other exploratory statistical procedures are neither of these, Baird argues, by virtue of their being merely information-transforming operations that only produce data-not knowledge and not hypotheses-that may or may not be useful. Baird then portrays me as arguing that factor analysis produces hypotheses which then must be interpreted. But what is to be interpreted, he counters is data, not hypotheses. And the process of interpretation, he argues, is extrinsic to the process of factor analysis. There may be little consensus about how to understand or interpret the factors [produced by a factor analysis]. But, there will be complete agreement about what the output is; no one will argue about what the specific linear equations