I recently uploaded a new version of the lme contributed library to
the main CRAN site. It should be available on mirrors within the next
24 hours.
This version matches the beta version of NLME 3.0 for S-PLUS. It only
provides methods for linear mixed-effects models in R so it is called
lme rather than nlme. The other big thing that is missing in the R
version is all the graphics because they rely on trellis graphics
functions.
Even with these things missing it is a big step forward for Jose'
Pinheiro and me to have this running under R. This version can handle
a substantial portion of the models fit by SAS PROC MIXED. It
computes either maximum likelihood (ML) or restricted maximum
likelihood (REML) estimates for linear mixed-effects models with
single or multiple nested levels of random effects.
The library includes a SAS_Mixed directory with R transcripts of
analyses paralleling those in the book "SAS System for Mixed Models"
by Littell, Milliken, Stroup and Wolfinger.
Please ensure that you get the lme_2.9-2.tar.gz file. The
lme_2.9-1.tar.gz file has problems with installation of the help files
and has the wrong version listed in the DESCRIPTION file. (If you
really want to install lme_2.9-1.tar.gz you can do so by removing the
file lme/man/[.pdMat.Rd before trying to run R INSTALL.)
This is still work in progress. It seems right now that the
variance-covariance matrices with compound symmetry structure are not
behaving properly but I haven't tracked down the problem there yet.
The speed is quite good and can be improved when we get the hooks into
the underlying optimization code in R. Here is a speed comparison
used in the "SAS Sytems for Mixed Models" book.
> library( lme )
> data( SIMS )
> ### Analysis of the Second International Mathematics Study (SIMS)
> ### described in section 7.2 of "SAS System for Mixed Models"
> unix.time(assign("fm1RSIMS", lme(Gain ~ Pretot, SIMS, ~ Pretot |
Class,
+ REML = TRUE)))
[1] 14.25 0.10 15.00 0.00 0.00> ### Timing was done on a Pentium Pro 200 MHz system running Linux
> ### Your mileage may vary.
> summary(fm1RSIMS) # compare to output 7.4, p. 262, "SAS
System ..."
Linear mixed-effects model fit by REML
Data: SIMS
AIC BIC logLik
22392.57 22429.85 -11190.29
Random effects:
Formula: ~Pretot | Class
Structure: General positive-definite
StdDev Corr
(Intercept) 3.8065218 (Intr)
Pretot 0.0959321 -0.641
Residual 4.7154814
Fixed effects: Gain ~ Pretot
Value Std.Error z-value p-value
(Intercept) 7.059605 0.365898 19.29390 0
Pretot -0.186031 0.016098 -11.55592 0
Correlation:
(Intr)
Pretot -0.76
Standardized Within-Group Residuals:
Min Q1 Med Q3 Max
-5.377621209 -0.595534211 0.009876494 0.640990584 3.718249985
Number of Observations: 3691
Number of Groups: 190
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
