christian <[EMAIL PROTECTED]> writes:
> Hello,
>
> ...have anybody experience or a good idea
> how i can display the result from the correspondence analysis
> in a better way like spss (i.e. excel ), because the visual performance
> in my humble opinion is not the best !?
I followed with interest the discussion about correspondence analysis
and comments by professor Nishisato. As it has been noticed
correspondence analysis has been very popular in France in the
seventies and specially in Paris where was Jean-Paul Benzecri.
Michael Greenacre who studied in Paris with Benzecri played a
an important role to make correspondance analysis known outside of
french speaking countries, but ther was also in 1984 the book:
"Multivariate Descriptive Statistical Analysis. Correspondence
Analysis and Related Techniques for Large Matrices". Wiley, 1984,
pp230. : Alain MORINEAU, Ludovic LEBART and Ken WARWICK.
The Leiden University Data Theory Scaling System Group lead by Jan De Leeuw
played a very important role to make correspondence analysis popular in
the early eighties. GIFI was an emanation of this group and they created
the categories software which became part of SPSS
French packages, for example SPAD which has been developped by
Lebart and Norineau (and others), have nice graphical output
for correpondence analysis and it is automatic. I use SPAD
to perform correspondance analysis but it is in french:-)
It is possible to use tricks to have nice plots with SPSS
but it is really complicate, you have to to take coordinates
produced by correspondence analysis and create a new data set
with them and then you can use scatter plot.
In french correspondence analysis is usually named "analyse
FACTORIELLE des correspondances" and people use the acronyms AFC and
AFCM (for multiple CA). In 1984-1986 I was involved in research
project in England (Lancaster), the aim of this project was to study
the difference between french statistical approaches in social
sciences based on correspondence analysis and what was done elsewhere
(english speaking area). I did a thesis on generalised linear modeling
and I did not know very well multivariate methods; I knew much better
classical statistics than correspondence analysis and it is England
that I learned how to use this method with an australian boss.
I consider that rotation (varimax or other) is quite useful
in factor analysis and I think rotation could be an improvement
of correspondence analysis.
--
Joseph Saint Pierre
http://www.cict.fr/cict/personnel/stpierre
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
