R help - Mar 2004 - Lattice, skip= and layout= problem, plotting object from nlme output

I generate a groupedData object

library(nlme)
obj <- groupedData(mg10 ~ time | gp, data = common, outer = ~pct)

gp has 101 levels, and pct has 3. There are 38, 25, 38 gps in each of the 
levels of pct respectively.

I fit my model

fit.rtg <- lme(mg10 ~ time * group,
    data = obj,
    random = ~time * group | gp)

Now I try to plot the results. I would like to print 40 panels on each page 
and have a new level of pct start a new page. The effect of using outer in 
the call of groupedData is that the values of gp are presented by pct.

plot.rtg <- plot(augPred(fit.rtg),
    skip = c(rep(FALSE, 38), TRUE, TRUE,
             rep(FALSE, 25), rep(TRUE, 15),
             rep(FALSE, 38), TRUE, TRUE),
    layout = c(5, 8, 3),
    strip = FALSE
)

What I get is 38 panels on page 1 with 2 blank panels top right. I had 
hoped to get 25 on the next page with 15 blanks but I get 38 again with 2 
blanks.

If I alter layout to (say) layout = c(12, 10) I get blanks as expected on 
the single page produced so I surmise that skip is forcing the same format 
on each page. There is an example in \cite{pinheiro00} which suggests that 
what I wanted can be done (p113,  p445-447) so I suspect I am doing 
something stupid here.

I am using R 1.8.1 with the versions of lattice and nlme that came with it. 
I am using Windows 98SE if that is relevant.

@BOOK{pinheiro00,
   author = {Pinheiro, J C and Bates, D M},
   year = 2000,
   title = {Mixed-effects models in {S} and {S-PLUS}},
   publisher = {Springer-Verlag}
}


Michael Dewey
m.dewey at iop.kcl.ac.uk

On Monday 22 March 2004 07:15, Michael Dewey wrote:
> I generate a groupedData object
>
> library(nlme)
> obj <- groupedData(mg10 ~ time | gp, data = common, outer = ~pct)
>
> gp has 101 levels, and pct has 3. There are 38, 25, 38 gps in each of
> the levels of pct respectively.
>
> I fit my model
>
> fit.rtg <- lme(mg10 ~ time * group,
>     data = obj,
>     random = ~time * group | gp)
>
> Now I try to plot the results. I would like to print 40 panels on
> each page and have a new level of pct start a new page. The effect of
> using outer in the call of groupedData is that the values of gp are
> presented by pct.
>
> plot.rtg <- plot(augPred(fit.rtg),
>     skip = c(rep(FALSE, 38), TRUE, TRUE,
>              rep(FALSE, 25), rep(TRUE, 15),
>              rep(FALSE, 38), TRUE, TRUE),
>     layout = c(5, 8, 3),
>     strip = FALSE
> )
>
> What I get is 38 panels on page 1 with 2 blank panels top right. I
> had hoped to get 25 on the next page with 15 blanks but I get 38
> again with 2 blanks.
>
> If I alter layout to (say) layout = c(12, 10) I get blanks as
> expected on the single page produced so I surmise that skip is
> forcing the same format on each page. There is an example in
> \cite{pinheiro00} which suggests that what I wanted can be done
> (p113,  p445-447) so I suspect I am doing something stupid here.
You are seeing documented lattice behavior, but looks like that's 
inconsistent with what happens in S-PLUS. lattice replicates the skip 
vector to be as long as the number of panels per page and uses that for 
every page, while S-PLUS replicates it to be as long as the total 
number of panels spanning all pages.

I can easily fix this for 1.9.0, but this may break old (but probably 
rare) lattice code. For example, 

layout = c(2,2,2), skip = c(T,F,F)

will now expand to 

page 1: T, F, F, T
page 2: F, F, T, F

as opposed to the current behavior

page 1: T, F, F, T
page 2: T, F, F, T

Any objections to that ? 

Deepayan

At 13:28 22/03/04 -0600, you wrote:

[snip my original problem]

>You are seeing documented lattice behavior, but looks like that's
>inconsistent with what happens in S-PLUS. lattice replicates the skip
>vector to be as long as the number of panels per page and uses that for
>every page, while S-PLUS replicates it to be as long as the total
>number of panels spanning all pages.
>
>I can easily fix this for 1.9.0, but this may break old (but probably
>rare) lattice code. For example,
>
>layout = c(2,2,2), skip = c(T,F,F)
>
>will now expand to
>
>page 1: T, F, F, T
>page 2: F, F, T, F
>
>as opposed to the current behavior
>
>page 1: T, F, F, T
>page 2: T, F, F, T
>
>Any objections to that ?
 From my point of view the current behaviour is not very useful, and in 
your example either behaviour seems a bit strange so I would not object to 
it changing.

Thanks for the quick response, and thanks for making lattice available. I 
find myself making so many more graphical displays because they are (a) 
easy (b) aesthetically pleasing (c) revealing about my dataset.
>Deepayan
Michael Dewey
m.dewey at iop.kcl.ac.uk