-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org]
On Behalf Of Deepayan Sarkar
Sent: January-12-11 7:26 AM
To: Joshua Wiley
Cc: r help
Subject: Re: [R] how to change strip text of effect plot
On Wed, Jan 12, 2011 at 10:36 AM, Joshua Wiley <jwiley.psych at gmail.com>
wrote:
Hmm, I felt like it was the clearest, most direct way. Then again,
things of this nature (overide defaults using arguments rather than
changing what you feed the functions) do seem to be a common request
both for lattice and ggplot2. In any case, my memory of the lattice
book and a quick search of the archives suggest that the typical way
to do this when dealing with lattice functions such as xyplot() would
be to define a custom strip function. For example:
strip = function(..., factor.levels) strip.default(..., factor.levels
= c("low", "medium", "high"))
however, if I am not mistaken, you are plotting an object of class
"efflist", which dispatches plot.efflist which eventually dispatches
plot.eff, in the bowels of plot.eff is the call to xyplot(), with its
own strip function already defined. I did not see a way to override
this, nor did I see an option to pass an argument to it.
It is trivial to edit the function's code to work with such a change.
You could probably create a local copy of it, make the necessary
changes, and then just ensure that your object gets dispatched to your
revised function rather than the packages original. Even easier, I
suppose:
debug(plot)
plot(eff.cowles, 'neuroticism:ex2',factor.names=F) ## follow along
until it has just created the object "plot"
## then overwrite that with your desired changes (factor.levels) ##
before it is printed to the screen
plot <- xyplot(eval(parse(text = paste("fit ~", predictors[x.var],
"|", paste(predictors[-x.var], collapse = "*")))), strip =
function(..., factor.levels) strip.default(..., factor.levels =
c("low", "medium", "high"),
strip.names = c(factor.names, TRUE)), panel = function(x,
y, subscripts, x.vals, rug, lower, upper, has.se, ...) {
llines(x, y, lwd = 2, col = colors[1], ...)
if (rug)
lrug(x.vals)
if (has.se) {
llines(x, lower[subscripts], lty = 2, col = colors[2])
llines(x, upper[subscripts], lty = 2, col = colors[2])
}
if (has.thresholds) {
panel.abline(h = thresholds, lty = 3)
panel.text(rep(current.panel.limits()$xlim[1],
length(thresholds)),
thresholds, threshold.labels, adj = c(0, 0),
cex = 0.75)
panel.text(rep(current.panel.limits()$xlim[2],
length(thresholds)),
thresholds, threshold.labels, adj = c(1, 0),
cex = 0.75)
}
}, ylim = ylim, ylab = ylab, xlab = if (missing(xlab))
predictors[x.var]
else xlab, x.vals = x.vals, rug = rug, main = main, lower =
x$lower,
upper = x$upper, has.se = has.se, data = x, scales = list(y =
list(at = tickmarks$at,
labels = tickmarks$labels), alternating = alternating),
...)
alternately still, you could rename the factor levels in the object
"eff.cowles", which is sort of inbetween changing your data and
changing the strip label defaults.
I can understand why it seems like there should be a simpler solution,
but I honestly do not see one. The factor.levels argument does not
even work inside xyplot(), it needs to be in the strip function, and
that is nested far away from plot(effobject). I do not see any
documentation that suggests a way, nor do I see any formal arguments
to facilitate it. Then again, I'm not an expert in lattice or the
effects package.
One generally useful fact is that if you can get hold of the "trellis"
object being plotted, you may be able to change the strip labels before
plotting it. For example,
foo <- trellis.last.object()
foo <- plot(eff.cowles[['neuroticism:ex2']],factor.names=F)
(this returns the object, but also forces a plot), followed by
$ex2
[1] "(1.98,8.99]" "(8.99,16]" "(16,23]"
dimnames(foo)$ex2 <- c("low", "medium", "high") foo
But I would tend to agree with Josh, that specifying labels in the
cut() call is the more natural approach. It is not the job of graphics
functions to modify (characteristics of) your data.
-Deepayan
Best regards,
Josh
On Tue, Jan 11, 2011 at 8:06 PM, Wincent <ronggui.huang at gmail.com>
Sure that is one way to go.
Since it is possible to pass lattice arguments to that high level
function, and there should be ways to relabel/custom the strip text
in the lattice plotting system, I would think such an indirect method
a last resort.
Thank you.
Ronggui
On 12 January 2011 11:51, Joshua Wiley <jwiley.psych at gmail.com>
Hi,
I am guessing this is not what you meant by "on the fly", but I
think it will be by far the easiest way. Plotting an effects object
is a high level plot with a lot of defaults and automation built in
to make your life simple. The cost is that it is less
flexible---you work its way, not vice versa. If you want the factor
named high, just label it that way to begin with. If you think it
makes the graphs more interpretable/meaningful, then it will make
better also. Worst case, you fit the model twice (one with fancy
names, one with numbers), which unless you have a massive dataset
will not take long or be an onerous burden anyways. Here's how you
can include labels directly in cut():
Cowles$ex2 <- cut(Cowles$extraversion, 3, c("low", "medium",
"high")) mod.cowles <- glm(volunteer ~
sex+neuroticism*ex2,data=Cowles, family=binomial) eff.cowles <-
allEffects(mod.cowles) plot(eff.cowles,
'neuroticism:ex2',factor.names=F)
Cheers,
Josh
On Tue, Jan 11, 2011 at 7:21 PM, Wincent <ronggui.huang at gmail.com>
Dear r heper,
How can I change the strip text, for example (16,23] in the
following example, to other more informative text such as "high
level" on the fly?
library(effects)
Cowles$ex2 <- cut(Cowles$extraversion,3) mod.cowles <-
glm(volunteer ~ sex+neuroticism*ex2,data=Cowles, family=binomial)
eff.cowles <- allEffects(mod.cowles) plot(eff.cowles,
'neuroticism:ex2',factor.names=F)
Thank you.
Ronggui