Factor scores have a
long tradition within many disciplines, and indeed at one time were
very progressive. In some disciplines, they might still be important
tools. Thus, we provided this brief chapter. However, we also urge
caution in their use. Given what we know about the volatility of EFA
analyses (even when the scale is traditionally strong with an unusually
clear factor structure), and the conceptual and mathematical issues
with factor scores (e.g., controversial mathematical issues like indeterminacy
dating back to the early 20th century), individuals wanting to use
this technique must be aware of these limitations. It might be useful
to attempt to estimate the extent to which factor loadings are volatile
or stable (particularly across samples or subgroups), and it might
also be useful to consider latent variable modeling (SEM) or modern
measurement alternatives (e.g., Rasch, IRT) instead. Rasch analysis,
IRT, and SEM have not solved all our problems in measurement, but
they might have advantages for some applications. (See also DiStefano,
Zhu, & Mindrila (2009.) They are better suited to large samples,
but as we hope we have persuaded you by now, EFA should also be a
large sample procedure.