ABSTRACT
As we indicated in Chapter 6, factor analysis is a hypothesis-developing method to be used principally at the outset of a research program designed to suggest hypothetical variables that may be immediate, common causes of a set of observed variables. The assumption is that the common causes are “latent” and not directly measured, which is why they are hypothetical. Another assumption is that the common causes are fewer in number than the observed variables studied. As explanatory entities, the common factors need to be fewer because there would be no saving in thought if the number of explanatory entities equaled the number of entities to be explained (Thurstone, 1937). Furthermore, we assume that the common factors (as we will call the latent variables) are not dependent on the particular set of observed variables for their existence, but rather that they exist in their own right, and may be found in studies with other observed variables. Furthermore, “The factorial description of a test must remain invariant when the test is moved from one battery to another” (Thurstone, 1937, p. 75). This implies that a variable’s factor pattern coef cients should be invariant regardless of what other variables are embedded with it. The indeterminacy of the common and unique factors of the common-factor model, which mathematically results from there being only n observed variables with which to determine n + r latent variables, allows the latent variables to be more than what may be known of them from the observed variables.
