a very preliminary lme for R

Douglas Bates bates@stat.wisc.edu
Tue, 17 Feb 1998 16:17:17 -0600 (CST)


Jose and I have a version of the lme (linear mixed effects) library
that is known to work with at least one example on at least one
version of R on at least one computer so that means it is time to
upload to CRAN, right?

In the near future a file lme_2.8-0.tar.gz will appear in the
src/contrib/devel area under CRAN.  Please remember this is a snapshot
of the development version.  We know there are things that don't work
with it under R (like all the trellis-based graphics, for example) but
it does currently work in R to the extent that it will fit
mixed-effects models using maximum likelihood or restricted maximum
likelihood (REML), including nested models.

The documentation is, shall we say, "sketchy".  It is more than
non-existent but not much more.  In other words this release is for
the brave early adopter who doesn't mind reading the source code to
figure out how the damn thing works.

If you do pick it up you may find it helpful to look at some recent
presentations we have given on the design of the library.  These are
available at
		 http://franz.stat.wisc.edu/pub/NLME/
right beside one of the CRAN archive sites
		 http://franz.stat.wisc.edu/pub/CRAN/
A paper on the computational methods used in this version of the
library will appear there "real soon".

Features of the library include:
 - use of a "groupedData" class that extends the data.frame class to
   describe relationships amoung the variables and the grouping of the
   observations. 
 - multiple levels of grouping in the data and nested random effects
   in the models.
 - summarization of data according to grouping of the observations via 
   the gsummary function (don't know if this works in R yet).
 - trellis-based plots of data and fitted models and derived
   quantities (definitely doesn't work in R yet)
 - much more stable computational methods, selection of starting
   values, etc. for fitting linear mixed-effects models
 - maximum likelihood or restricted maximum likelihood estimation
   criteria

Our interest in putting this up is to get feedback.  Please be gentle.

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel 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-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._