plot with nlme

From: Hank Stevens <HStevens at muohio.edu>
To: r-help at stat.math.ethz.ch
Subject: [R] plot with nlme
Date: Tue, 29 Apr 2003 10:03:23 -0400

Using R v. 1.7.0 on Windows 2000

I would like to plot the fitted values of a model as a function of a 
continuous covariate, augmented with data (e.g., augPred) grouping by 
combinations of fixed effects. I have not been able to use augPred 
effectively, and am wondering if it does not handle unbalanced data (3 
out of 192 missing).
I include below the model and an xyplot that almost does the job. I would 
happily send anyone the data if they would be willing to help.
Many Thanks in advance.
Hank

The model:

fm <- lme( log(S,2) ~ Fert*Litter*Seed*Density, data = collimd, random = 
~1|block/plot/Seed )

# The following is close, but fits lines within each panel rather than 
giving me the fitted values generated by the model.

xyplot( log(S,2)  ~  log10(Nper0.5m)  |  Fert*Seed, data = collimd,
groups=Litter,
       scales = list( alternating =FALSE, tck=c(1,0)), layout = 
c(4,2,1), aspect = 1, as.table = T,
         key = list(space="top", transparent = TRUE,
                   points=list(pch=c(1,3)),
                     #trellis.par.get("superpose.symbol")$pch[1:2],
                   lines=list(lty=1:2),
                   text = list(c("With Litter", "Litter Removed"))   ) ,

       panel = function(x, y, subscripts,...) {
           panel.grid(h=-1, v= -1)
           panel.superpose(x, y,subscripts, pch=c(1,3),col=1,...)
            panel.superpose(x, y, 
panel.groups="panel.lmline",lty=1:2,subscripts,...)

           })



Martin Henry H. Stevens, Assistant Professor
338 Pearson Hall
Botany Department
Miami University
Oxford, OH 45056

Office: (513) 529-4206
Lab: (513) 529-4262
FAX: (513) 529-4243
http://www.cas.muohio.edu/botany/bot/henry.html
http://www.muohio.edu/ecology

Martin Henry H. Stevens, Assistant Professor
338 Pearson Hall
Botany Department
Miami University
Oxford, OH 45056

Office: (513) 529-4206
Lab: (513) 529-4262
FAX: (513) 529-4243
http://www.cas.muohio.edu/botany/bot/henry.html
http://www.muohio.edu/ecology

At 10:54 AM 4/29/2003, Francisco Zagmutt wrote:

Hi there

I had that problem before and apparently it has to be with the log 
transformation that you used in the formula. R will not recognize the 
variable name and hence will just give you a plot with the fitted values 
and will give you a warning about the variable name.
Try creating a new variable with the log transformed values (i.e. log.y) 
of your original predictor and then use this new variable as your 
predictor.

fm <- lme( log.y ~ Fert*Litter*Seed*Density, data = collimd,

random = ~1|block/plot/Seed )

Let me know if this worked

Francisco

From: Hank Stevens <HStevens at muohio.edu>
To: "Francisco J. Zagmutt Vergara" <fzagmutt at hotmail.com>
Subject: Re: [R] plot with nlme
Date: Tue, 29 Apr 2003 11:06:27 -0400

Francisco!
Thanks!

plot( augPred( fm) )

worked!

Now, however, I would like to group by fixed effects, not random variables 
(e.g., Subjects). Any thoughts?