[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: coefficients in factor analysis
Irina Paley <Irina.Paley@occ.treas.gov> asked about the
difference between "factor loadings", as shown in the output of
the -factor- command, and the "scoring coefficients", as shown
when using -predict- after -factor-.
Nick Cox <firstname.lastname@example.org> replied back with some guidance
on the mystical aspects of factor terminology and suggested that
Irina consult the Stata manuals and text books on the subject.
In addition to the text books found in the reference section of
the "[MV] factor" manual entry, you might find the chapter on
factor analysis from
Tabachnick, Barbara G. and Linda S. Fidell. 2007. "Using
Multivariate Statistics, 5ed". Boston:Allyn and Bacon.
more approachable than some of the more mathematically complete
references that are available. If you can't find the 5th
edition, which just came out, earlier editions will be adequate
for learning the basics. A downside of the book is that they use
a couple of other packages instead of Stata for showing their
Just as a quick summary answer for your question -- the factor
loadings are the correlations between your observed variables and
the unknown factors you are trying to estimate.
The factor scoring coefficients are the coefficients to use with
the standardized versions of your variables in order to obtain
the factor scores for your observations.
See pages 242-247 in "[MV] factor postestimation" in the Version
9 Stata manuals for examples and discussion of factor scores.
Ken Higbee email@example.com
* For searches and help try: