1: Two techniques, principal axes
factor extraction and iterated principal axes factor extraction, are
PCA methods and not EFA by default. Initial extraction of communalities
using the squared multiple correlation must be specified using the
PRIORS = SMC option in order for EFA extraction to occur.
[return]
2: Note the following detail where
an orthogonal rotation is used. Oblique rotations are slightly more
complicated if doing the calculation by hand, as the factors are correlated.
Communalities in this case are the sum of each variable’s pattern
loading multiplied by the structure loading. Statistical software
handles these complexities for us.
[return]
3: Please
note that alpha extraction and maximum likelihood extraction are weighted
extraction methods. The unweighted initial estimates are identical
but SAS presents only the weighted version of the initial estimates.
[return]
4: Interested
readers can get much more information on this and other technical
details of extraction in Nunnally & Bernstein (1994), particularly
Chapter 11. For our purposes, we aim to provide a more applied overview,
and so will skip most of the technical details.
[return]
5: Please note, there is one other
extraction method, principal components analysis. There are also two
additional extraction options to recognize and perform analyses on
a set of scoring coefficients or a factor pattern matrix. These are
excluded from this explanation because they are not specific EFA methods.
[return]