Hello S+ Users,
I'm using factanal in S+ version 7.0 on Windows
I find my out of sample fits have room for improvement, so I'd like to
improve them in a justifiable way.
In more detail, my general setup looks like this:
factanal(InputData, NumFactors, method="mle")
# InputData is typically 100 observations x 10 variables
# NumFactors is estimated to be from 1-4 depending on the data
set
Selection of NumFactors is done using the standard chi-squared
distributed test statistic provided with factanal, and done until p>0.05
for adding another factor to the model (usually the number of factors is
very obvious - p very small before the last, but p>0.05 after the last).
If I test the resulting NumFactors using N-fold cross-validation (hold
out an observation, estimate model on the other N-1, then apply it to
the held out observation, then repeat for remaining observations), my
fit is not nearly as good out of sample. For example, in one set of
cases where the test statistic says to fit 4 factors, I can perform
approximately as well out of sample using a simple 2 parameter model
where I estimate one average variance and one average correlation and
apply these estimates to every variable.
I can improve these out of sample fits using an heuristic form of
shrinkage akin to adjusted RSQ, but my justification for this is just
that, heuristic.
My QUESTION is: can anyone refer me to a reference that performs some
sort of shrinkage on factanal loadings in a way that improves the out of
sample fit, or some other method that allows me to do this? I would
have to guess that others using this technique would face similar
problems, but of course don't know this for sure.
Thank you very much,
Matt Kurbat
--
This message and any attachments are confidential, proprietary, and may be
privileged. If this message was misdirected, Barclays Global Investors (BGI)
does not waive any confidentiality or privilege. If you are not the intended
recipient, please notify us immediately and destroy the message without
disclosing its contents to anyone. Any distribution, use or copying of this
e-mail or the information it contains by other than an intended recipient is
unauthorized. The views and opinions expressed in this e-mail message are the
author's own and may not reflect the views and opinions of BGI, unless the
author is authorized by BGI to express such views or opinions on its behalf.
All email sent to or from this address is subject to electronic storage and
review by BGI. Although BGI operates anti-virus programs, it does not accept
responsibility for any damage whatsoever caused by viruses being passed.
|