R help - Jan 2015 - panel.xyplot and panel.loess using two different groupings

Hello,

I have a dataset consisting of four variables: species (factor, five
levels), C3C4 (factor, two levels), and numeric variables PAR and LRU.
I wish to produce a scatter plot of PAR vs LRU where (1) each species
has a unique symbol and color (2) there is an overlaid loess line that
is calculated for each C3C4 level.  The following code produces a nice
plot where each species has its own loess curve, but I'm struggling for
the syntax to produce a single loess curve for all the points in each of
the two plots:

xy <- xyplot(LRU~PAR|C3C4, data=light_data, groups=species,
             panel=panel.superpose,
             auto.key=TRUE,
             col.line=pal,
             par.settings=mytheme,
             panel.groups=function(x, y, ...){
                 panel.xyplot(x, y, ...)
                 panel.loess(x, y, ...)})

Any help much appreciated.  Thanks!  
Tim

--

Timothy W. Hilton
Assistant Project Scientist
Sierra Nevada Research Institute
University of California, Merced
thilton at ucmerced.edu

Learn to use custom panel functions to give you the flexibility and
features you require.

In this case, you want something like:

xyplot(LRU~PAR|C3C4, groups = species,
       panel= function(x,y,...){  ## custom panel function to add an
overall loess line
         panel.superpose(x,y,...)
         panel.loess(x,y,col="darkblue",...)
       },
       panel.groups = function(x,y,...){
         panel.xyplot(x,y,...)
         panel.loess(x,y,...)
       }
## plus any other key and plot options
       )


Note that this could fail if groups have too few points to fit a loess
curve, so you might want to add checks in the panel.groups function
for this.


HTH.


-- Bert

Bert Gunter
Genentech Nonclinical Biostatistics
(650) 467-7374

"Data is not information. Information is not knowledge. And knowledge
is certainly not wisdom."
Clifford Stoll




On Tue, Jan 27, 2015 at 10:19 AM, Timothy W. Hilton
<thilton at ucmerced.edu> wrote:
> Hello,
>
> I have a dataset consisting of four variables: species (factor, five
> levels), C3C4 (factor, two levels), and numeric variables PAR and LRU.
> I wish to produce a scatter plot of PAR vs LRU where (1) each species
> has a unique symbol and color (2) there is an overlaid loess line that
> is calculated for each C3C4 level.  The following code produces a nice
> plot where each species has its own loess curve, but I'm struggling for
> the syntax to produce a single loess curve for all the points in each of
> the two plots:
>
> xy <- xyplot(LRU~PAR|C3C4, data=light_data, groups=species,
>              panel=panel.superpose,
>              auto.key=TRUE,
>              col.line=pal,
>              par.settings=mytheme,
>              panel.groups=function(x, y, ...){
>                  panel.xyplot(x, y, ...)
>                  panel.loess(x, y, ...)})
>
> Any help much appreciated.  Thanks!
> Tim
>
> --
>
> Timothy W. Hilton
> Assistant Project Scientist
> Sierra Nevada Research Institute
> University of California, Merced
> thilton at ucmerced.edu
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

Thanks, Bert, for the reply.  This is very helpful.  I have to admit
I've read the docs for the panel and panel.groups arguments before and
gotten myself pretty confused.  Your small example is very helpful.

One more question for the list...

Bert's panel.loess col='darkblue' argument is being overridden by my
call to par.settings=mytheme, which I've defined like this (using the
RColorBrewer library):

pal <- brewer.pal(length(levels(light_data[['species']])),
'Dark2')
mytheme <- standard.theme("pdf")
mytheme[['superpose.symbol']][['pch']] <- c(15,16,17,3,4)
mytheme[['superpose.symbol']][['col']] <- pal

I've not been able to figure out which trellis.par.get() setting
controls the loess line colors, so I've not been able to set those in
the theme.  Similarly, I've not been able to pass a line color argument
to loess that doesn't get clobbered by the theme.  Is this possible?

Thanks,
-Tim

On Tue, Jan 2015, 27 at 11:13:51AM -0800, Bert Gunter
wrote:
> Learn to use custom panel functions to give you the flexibility and
> features you require.
> 
> In this case, you want something like:
> 
> xyplot(LRU~PAR|C3C4, groups = species,
>        panel= function(x,y,...){  ## custom panel function to add an
> overall loess line
>          panel.superpose(x,y,...)
>          panel.loess(x,y,col="darkblue",...)
>        },
>        panel.groups = function(x,y,...){
>          panel.xyplot(x,y,...)
>          panel.loess(x,y,...)
>        }
> ## plus any other key and plot options
>        )
> 
> 
> Note that this could fail if groups have too few points to fit a loess
> curve, so you might want to add checks in the panel.groups function
> for this.
> 
> 
> HTH.
> 
> 
> -- Bert
> 
> Bert Gunter
> Genentech Nonclinical Biostatistics
> (650) 467-7374
> 
> "Data is not information. Information is not knowledge. And knowledge
> is certainly not wisdom."
> Clifford Stoll
> 
> 
> 
> 
> On Tue, Jan 27, 2015 at 10:19 AM, Timothy W. Hilton
> <thilton at ucmerced.edu> wrote:
> > Hello,
> >
> > I have a dataset consisting of four variables: species (factor, five
> > levels), C3C4 (factor, two levels), and numeric variables PAR and LRU.
> > I wish to produce a scatter plot of PAR vs LRU where (1) each species
> > has a unique symbol and color (2) there is an overlaid loess line that
> > is calculated for each C3C4 level.  The following code produces a nice
> > plot where each species has its own loess curve, but I'm
struggling for
> > the syntax to produce a single loess curve for all the points in each
of
> > the two plots:
> >
> > xy <- xyplot(LRU~PAR|C3C4, data=light_data, groups=species,
> >              panel=panel.superpose,
> >              auto.key=TRUE,
> >              col.line=pal,
> >              par.settings=mytheme,
> >              panel.groups=function(x, y, ...){
> >                  panel.xyplot(x, y, ...)
> >                  panel.loess(x, y, ...)})
> >
> > Any help much appreciated.  Thanks!
> > Tim
> >
> > --
> >
> > Timothy W. Hilton
> > Assistant Project Scientist
> > Sierra Nevada Research Institute
> > University of California, Merced
> > thilton at ucmerced.edu
> >
> > ______________________________________________
> > R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> > https://stat.ethz.ch/mailman/listinfo/r-help
> > PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> > and provide commented, minimal, self-contained, reproducible code.