[R] Factor Analysis

Prof Brian Ripley ripley at stats.ox.ac.uk
Tue May 1 22:42:20 CEST 2001

On Tue, 1 May 2001, Magill, Brett wrote:

> Thanks to Brian Ripley for the time and effort put into developing the
> factanal package for R.  I have found it very useful.  However, in trying to
> replicate some results from previous research, I have run into the need for
> a couple of extensions and was wondering if they might find their way into
> future implementations.
> 1.  Though maximum likelihood estimates might be more rigorous and
> appropriate as a default, it would be great to have the option of conducting
> factor analysis based on  principle factors estimation.

principal?   I don't write code for methods I believe should come with a
severe health warning, but `patches may be accepted'.

> 2.  When a Heywood case is encountered, error out saying such.  Provide an
> option to perform the analysis anyway setting values to small positive
> numbers (although it might be statistically suspect, replicating results
> from other studies suggests to me that this is exactly what was done).

Unfortunately the precise value of the `small positive value' can be
critical.  However, I don't get this: this is exactly what factanal does
now, using 0.005 as the value.  I don't trust the optimization enought to
really declare a Heywood case adn give an error (and that is a comment on
the likelihood surface, not the optimizer).

Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch

More information about the R-help mailing list