Adding text labels to lattice plots with multiple panels

Hi Peter,

Thanks for the solutions.  The only problem now is that I have
duplicate values in my labels that are removed by the duplicated
function.  Perhaps a better example of the labels vector would be:
lab<- c(1,2,4,4,6,6)

Your first version would maintain the second 4 (because they are in
different panels) but leave a blank for the second 6.
1 2 4 4 6
The second version removes the second 4 and 6.
1 2 4   6

The following works, but it is not very efficient for my full data set
with six levels of f1, two levels of f2, and 64 replicates of each of
these combinations.

bwplot(dv ~ f1 | f2, data = df, ylim = c(0.5, 1),
  panel = function(x, y, ..., subscripts) {
    lab<<- df$lab[subscripts]
    lab[c(2:3,5:6,8:9)]<<- ""
    panel.bwplot(x, y,  ...)
    panel.text(x, 0.55, labels = lab)
  }
)

Thanks,
Jeff


On Sat, Apr 9, 2011 at 7:52 PM, Peter Ehlers<ehlers at ucalgary.ca>  wrote:

On 2011-04-09 06:21, Dennis Murphy wrote:

Hi:

One hopes that there is a more elegant solution than this bit of
ad-hockery.

 From your posted example:

f1<- c(rep(c(rep("a", 3), rep("b", 3), rep("c", 3)), 2))
f2<- c(rep("A", 9), rep("B", 9))
dv<- c(0.9, 0.8, 0.85, 0.6, 0.65, 0.7, 0.8, 0.85, 0.8, 0.95, 0.85,
  0.9, 0.65, 0.7, 0.75, 0.85, 0.9, 0.85)
df<- data.frame(f1, f2, dv)
df$lab<- rep(1:6, each = 3)

df$lab2<- ''
df$lab2[seq(1, 16, by = 3)]<- 1:6        # adapt to your situation -
seq(1,
nrow(df) - 63, by = 64), perhaps

bwplot(dv ~ f1 | f2, data = df, ylim = c(0.5, 1),
  panel = function(x, y, ..., subscripts) {
    lab<- df$lab2[subscripts]
    panel.bwplot(x, y,  ...)
    panel.text(x, 0.55, labels = lab)
  }
)

Alternatively, panel.text() takes an alpha = argument; for example,

bwplot(dv ~ f1 | f2, data = df, ylim = c(0.5, 1),
  panel = function(x, y, ..., subscripts) {
    lab<- df$lab[subscripts]
    panel.bwplot(x, y,  ...)
    panel.text(x, 0.55, labels = lab, alpha = 0.5)
  }
)

You could toy with the value of alpha until something acceptable emerges.
But as I said, there is probably a better solution and I'm happy to be
educated if there is.

Here's a slight variation on your first solution which doesn't
require the data to be appropriately sorted, using your df:

  bwplot(dv ~ f1 | f2, data = df, ylim = c(0.5, 1),
  panel = function(x, y, ..., subscripts) {
    lab<- df$lab[subscripts]
    lab[duplicated(lab)]<- ""
    panel.bwplot(x, y,  ...)
    panel.text(x, 0.55, labels = lab)
  }
  )

and another variation which sets the text positions to NA for
all but the first pass through the panel.text() function:

  bwplot(dv ~ f1 | f2, data = df, ylim = c(0.5, 1),
  panel = function(x, y, ..., subscripts) {
    at.y<- rep(0.55, nrow(df))
    is.na(at.y)<- which(duplicated(df$lab))
    panel.bwplot(x, y,  ...)
    panel.text(x, at.y[subscripts], labels = df$lab[subscripts])
  }
  )

I think that the alpha argument is too one-off, i.e. dependent on
how many levels in the boxplot.

Peter Ehlers

HTH,
Dennis

On Sat, Apr 9, 2011 at 4:56 AM, Jeff
Stevens<stev0175 at googlemail.com>wrote:

Hi,

Thanks for the work around, Dennis.  My actual data set has 64
replicates for each factor level combination (rather than the 3 in the
example), so the overplotting is quite messy.  Any ideas on how to
avoid the overplotting?

Jeff


On Fri, Apr 8, 2011 at 7:32 PM, Dennis Murphy<djmuser at gmail.com>    wrote:

Hi:

After a number of false starts, I finally consulted Deepayan's book and

the

example on p. 73, suitably adapted, yielded a solution. Add a variable

for

the labels and then...

df$lab<- rep(1:6, each = 3)
bwplot(dv ~ f1 | f2, data = df, ylim = c(0.5, 1),
  panel = function(x, y, ..., subscripts) {
    lab<- df$lab[subscripts]
    panel.bwplot(x, y,  ...)
    panel.text(x, 0.55, labels = lab)
  }
)

If you look closely, you'll see that each label is overplotted three

times.

A similar plot in ggplot2  would be

library(ggplot2)
ggplot(df, aes(x = f1, y = dv)) + geom_boxplot() +
     geom_text(aes(x = as.numeric(f1), lab = lab), y = 0.55, alpha = 0.5)

+

     facet_wrap( ~ f2) + ylim(0.5, 1)

The alpha argument in geom_text() is designed to mitigate the

overplotting

effect somewhat.

HTH,
Dennis


On Fri, Apr 8, 2011 at 5:32 AM, Jeff Stevens<stev0175 at googlemail.com>
wrote:

Hi,

I am trying to add text to the bottom of a lattice bwplot with
multiple panels.  I would like to add a label below each boxplot, but
the labels do not come from the data.  I've tried the following, code:

f1<- c(rep(c(rep("a", 3), rep("b", 3), rep("c", 3)), 2))
f2<- c(rep("A", 9), rep("B", 9))
dv<- c(0.9, 0.8, 0.85, 0.6, 0.65, 0.7, 0.8, 0.85, 0.8, 0.95, 0.85,
0.9, 0.65, 0.7, 0.75, 0.85, 0.9, 0.85)
df<- data.frame(f1, f2, dv)
lab<- c(1, 2, 3, 4, 5, 6)

bwplot(dv ~ f1 | f2, data = df, ylim = c(0.5, 1),
  panel = function(x, y, ...) {
    panel.bwplot(x, y, ...)
    panel.text(x, 0.55, labels = lab)
  }
)

I have two problems.  First, the label values are writing over one
another.  I wrote a horrible hack (below) that fixes that problem, but
I would like to know if there are better/more flexible ways to do
this.

bwplot(dv ~ f1 | f2, data = df, ylim = c(0.5, 1),
  panel = function(x, y, ...) {
    x2<- as.factor(c("a", "b", "c"))
    panel.bwplot(x, y, ...)
    panel.text(x2, 0.55, labels = lab)
  }
)

Second, when using the horrible hack the first values are repeated in
the second panel.  How do I display the correct values in the second
panel?  The subscripts argument seems to do the trick when the labels
come from the data, but how do I get the proper values when the labels
come from outside of the data?


Many thanks,
Jeff

--
Jeff Stevens
Center for Adaptive Behavior and Cognition
Max Planck Institute for Human Development
Lentzealle 94
14195 Berlin, Germany

        [[alternative HTML version deleted]]