Although authors have
been presenting methods for summarizing replication in EFA for half
a century and more, most summarization techniques have been flawed
or less informative than ideal. In the 21st century, with CFA invariance
analysis as the gold standard for assessing generalizability and
replicability, replication within EFA has an important role to play—but
a different role than half a century ago. Today, replication in EFA
is a starting point—it adds value to EFA analyses in that it
helps indicate the extent to which these models are likely to generalize
to the next data set, and also helps to further identify volatile
or problematic items. This information is potentially helpful in the
process of developing and validating an instrument, as well as for
potential users of an instrument that has yet to undergo CFA invariance
analysis.
However, there are often
barriers to replication analysis. Foremost among these barriers is
the lack of adequate sample size in most EFAs that are reported in
the literature. The first priority for researchers should be adequate
samples. The second should be estimation of the replicability (or
stability) of the model presented. In the next chapter we review bootstrap
analysis as a potential solution to this issue, as it allows use of
a single, appropriately large sample to estimate the potential volatility
of a scale.