[R] mclust was Re: R or Splus

[markhall@gol.com]

> > However... I could become an R convert if I'm told it's
> >
> > a) mind-bogglingly simple to install on Win Nt
> >    there are firm limits to my geek credentials
> > b) can handle cluster analysis with 2000 records (mclust, preferred,
> >    although I'd take kmeans if nec'y).  Splus 5.0 on the
> >    Manchester Cray has been falling over on this, much to my great
> >    annoyance, no matter how large a workspace I give mclust, hclust
> >    or kmeans, etc.
> 
>   There is an mclust package on CRAN
> (<a href="http://cran.r-
project.org/src/contrib/PACKAGES.html#mclust),">http://cran.r-
project.org/src/contrib/PACKAGES.html#mclust),</a>I don't have
> any idea what the memory limitations are.  Do you know enough about the
> clustering algorithm (I don't) to know if this kind of analysis would
> involve mindbogglingly large matrices that would choke any system?
> 

Its hard to say from the information you give.  I would suggest you read the 
technical reports produced by Fraley and Raferty which are freely available 
from the Univ. of Washington Stats Dept. web site.  Most recently, is their 
Technical Report 380 entitled "Model-Based Clustering, Discriminant Analysis, 
and Density Estimation."  In part to answer your question, it isn't a matter or 
R versus S, but in terms of the number of variables and the model you are 
trying to fit.  Take a look at some of Fraley and Raferty's stuff.

Best, Mark Hall




-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._